Predicate Formation in the Verbal System of Modern Hebrew [Reprint 2010 ed.] 9783110882285, 9783110133073

Citation preview

Predicate Formation in the Verbal System of Modern Hebrew

Functional Grammar Series This series comprises monographs and collections written in the framework of Functional Grammar. The aim is to seek explanations for a wide variety of linguistic phenomena, both language specific and cross-linguistic, in terms of the conditions under which and the purposes for which language is used. Editors: A. Machtelt Bolkestein Simon C. Dik

Casper de Groot J. Lachlan Mackenzie General address: Institute for General Linguistics Functional Grammar Spuistraat210 NL-1012 VT Amsterdam The Netherlands Other books in this series: 1. A.M. Bolkestein, C. de Groot and J.L Mackenzie (eds.) Syntax and Pragmatics in Functional Grammar 2. A. M. Bolkestein, C. de Groot and J.L. Mackenzie (eds.) Predicates and Terms in Funtional Grammar 3. Michael Hannay English Existentials in Functional Grammar 4. Josine A. Lalleman Dutch Language Proficiency of Turkish Children born in the Netherlands 5. Jan Nuyts and Georges de Schutter (eds.) Getting one's Word into Line 6. Johan van der Auwera and Louis Goossens (eds.) Ins and Outs of the Predication

Other studies on Functional Grammar include S.C. Dik, Functional Grammar (1978), T. Hoekstra et al. (eds.). Perspectives on Functional Grammar (1981), S.C. Dik (ed.), Advances in Functional Grammar (1983). All published by FORIS PUBLICATIONS.

Predicate Formation in the Verbal System of Modern Hebrew

1987 FORIS PUBLICATIONS

Dordrecht - Holland/Providence Rl - USA

Published by: Foris Publications Holland P.O. Box 509 3300 AM Dordrecht, The Netherlands Sole distributor for the U.S.A. and Canada: Foris Publications USA, Inc. P.O. Box 5904 Providence Rl 02903 U.S.A.

Sole distributor for Japan: Sanseido Book Store, Ltd. 1-1, Kanda-jimbocho-cho Chiyoda-ku Tokyo 101, Japan CIP-DATA Junger, Judith Predicate Formation in the Verbal System of Modern Hebrew / Judith Junger. Dordrecht [etc.]: Foris. - (Functional Grammar Series; 7) Also published as a thesis University of Amsterdam, 1987. - With ref. ISBN 90-6765-368-3 SISO hebr-n 837 UDC 809.244-5 Subject heading: Modern Hebrew; functional grammar

ISBN 90 6765 368 3 © 1987 Foris Publications - Dordrecht No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission from the copyright owner. Printed in the Netherlands by ICG Printing, Dordrecht.

dr«V* to Cor

ERRATA

1. On page 10 end of paragraph one should read "The MH phonological system has 20 distinct consonant sounds — 2. On page 10 " /k/ - velar plosive" it should be added (has two othographic symbols). 3. On page 10 the last sentence of point 1. should read: "Thus no difference is made between the two orthographic signs for /t/, /v/, I si and /k/". 4. On page 172 the second name, "Kimhi, R.D." "Kimhi, D."

should read

5. On page 175 "Kimhi, R.D." should read: Kimhi, D. 6. On page 178 on the top HISTAFEL should read HIU>TAFEL, and in the middle "M FOAL·" should read: M9FOAL·. 7. On page 180 "SAFEL (SIFEL·)" should read: §AFEL (SlFEL), and "SUFAL" should read: &JFAL.

Contents

ACKNOWLEDGEMENTS

6

LIST OF ABBREVIATIONS AND SYMBOLS

7

INTRODUCTION

8

NOTES ON THE TRANSCRIPTION CHAPTER 1: LITERATURE SURVEY 1.0 Introduction 1.1 The traditional view 1.1.1 Gesenius 1.1.2 Kimhi 1.2 The 'lexicalist' view 1.2.1 Oman 1.2.2 Schwarzwald 1.3 Traditional and Transformational Generative views 1.3.1 Berman 1.3.2 Ariel 1.3.3 Sasson 1.4 Non-concatenative morphology 1.5 The FG view 1.5.1 In defense of the biayanim 1.5.2 Which biayania 1.6 Summary CHAPTER 2: RELEVANT ASPECTS OF FUNCTIONAL GRAMMAR

2.0 2.1 2.2 2.3

Introduction General outline of Functional Grammar The fund and the lexicon Basic terms and basic predicates 2.3.1 Derived predicates 2.4 Predicate formation 2.5 Applying the predicate frames to MH binyanim 2.6 Summary CHAPTER 3: THE LEXICON AND THE ROOTS

3.0 3.1 3.2 3.3

Introduction 'Dictionary* vs. 'lexicon' Approaches to the Hebrew lexicon The root

10 12 12 13 13 16 17 17 18 19 19 21 23 23 25 25 27 30 31

31 31 34 34 37 37 38 41 42

42 42 43 45

3.4 An FG lexicon for Modern Hebrew 3.4.1 The root 3.4.2 The patterns 3.4.2.1 The miSqalim 3.4.2.2 The binyanint 3.5 The organization of the lexicon 3.6 Summary

48 48 50 50 52 54 61

CHAPTER 4: PREDICATE FORMATION RULES: THE PRODUCTIVE PROCESSES IN THE BINYANIM SYSTEM

4.0 Introduction 4.1 Transitivity in Hebrew 4.1.1 Transitivity or valency 4.1.2 Productivity 4.2 Basic and derived binyanim 4.3 The basic binyanim 4.4 The derived biayaaiia 4.4.1 Patterns that can be both basic and derived 4.5 Valency increase 4.5.1 Causativity- a working definition 4.5.2 The semantic function of the causee 4.5.3 Derivation 4.6 Valency reduction 4.6.1 Valency reduction vs. Subject assignment to Go 4.6.2 Valency reduction in FG 4.6.2.1 First argument reduction: the NIFAL, PUAL and HUFAL patterns 4.6.3 Second argument reduction 4.6.4 Double reduction of valency 4.7 Summary CHAPTER 5 : THE CONFIGURATIONAL· SUB-SYSTEMS

> NIFAL -v

97 98 99 100 101 106 108 117 119 121 124 124

-v

5.5.1.2 PIEL

> PUAL; HIFIL > HUFAL -v 5.5.1.3 PAAL / PIEL· / HIFIL· > HITPAEL·

5.5.1.4 HITPAEL resulting from qualitative val. change 5.5.2 Valency increase +v 5.5.2.1 PAAL· —> HIFIL· +v 5.5.2.2 PAAL > PIEL·

83 88 92 95 97

5.0 Introduction 5.1 Theoretical background 5.2 The configurational sub-systems 5.2.1 The procedure for establishing the CSS's 5.2.2 Methodological remarks 5.2.3 General picture of the root-biayaaim distribution 5.3 The Configurational sub-systems 5.3.1 The size of the CSS's 5.4 Verification of the hypotheses 5.5 The derivational relations 5.5.1 Valency reduction -v 5.5.1.1 PAAL

62

62 64 64 70 70 71 72 73 73 74 76 80 82 82 83

127 131

140 141 141 144

5.6 Summary 5.6.1 Regarding the Modern Hebrew verbal system 5.6.2 Regarding Functional Grammar CHAPTER 6 : CONCLUSIONS

6.1 Results 6.1.1 The Hebrew verbal system 6.1.2 Functional Grammar 6.2 Future perspectives

148 148 152 153

153 153 155 157

MOTES

REFERENCES

168

AUTHOR INDEX

175

SUBJECT INDEX

176

Acknowledgements

This book, a study on the verbal system of Modern Hebrew and of the predicate formation mechanism of Functional Grammar, is the result of a three year research project financed by the Netherlands Organization for the Advancement of Pure Research (ZWO), under guidance of Prof. Dr. S.C. Dik (project number 300-163-019). This final result is due to the interest, encouragement and stimulation of many people. In the first place I wish to thank Prof. Dr. S.C. Dik for his insightful remarks and stimulating questions. My years of working with him have been both inspiring in doing research and educative in organising and writing down the results in a coherent manner. Many thanks are due also to Dr. A.M. Bolkestein and Dr. Y. Tobin, my copromotors, for their penetrating criticism and valuable suggestions, and not less for their involvement and encouragement. Thanks are due also to Prof. Dr. J.H. Hospers, Dr. Y. Ziv, Prof. Dr. K. Ehlich, Prof. 0. Schwarzwald and Prof. R. Berman for their helpful feedback on various parts of this research, as well as to the editors of the Functional Grammar series Dr. L. Mackenzie and Drs. C. de Groot. There are many others, colleagues and students, who helped me by discussing questions that came up, or by comments during lectures on various parts of this research both in the Netherlands and in Israel. I cannot list all of them here, but I wish to express my gratitude to them too, as well as to G. Watson for the thorough editing of the manuscript. I wish to thank my parents, who always stimulated my curiosity and enabled me to start on the academic path. And last but not least, thanks go out to Cor who by his humorous and sceptic remarks on linguistics in general, and this work in particular, made me have a more realistic perspective on my work. Gouda July 1987

Judith Junger

List of abbreviations and symbols

Semantic functions: Agent Ag Go = Goal Comp = Complement Rec - Recipient Ben - Beneficiary Exp - Experiencer Phen = Phenomenon Instr - Instrument Loc = Locative Temp = Temporal Po Positioner Proc = Processed Fo Force Dir = Direction 0 Zero function

Gender markers: m = masculine f = feminine

Syntactic functions: Subj = Subject Obj = Object

Bl B2 B3 B4 B5 B6 B7

Number markers: s = singular pi = plural Person markers Ip = first person 2p = second person 3p = third person tie binyanim:

Pragmatic functions Top = Topic Foe = Focus

= = = = = = =

PAAL MIFAL PIEL PUAL HIFIL HUFAL HITPAEL

Morpbo-syntactic categories: +tr = transitive -tr = intransitive pp - pseudo-passive i = inchoative ref = reflexive rec = reciprocal c = causative

Categories: V = verbal N = nominal A = adjectival

General: xl — xn = yl...yn = = IT* =

Aktionsart hab = habitual iter = iterative mom = momentaneous sud = sudden

arbitrary arguments arbitrary satellites arbitrary predicate arbitrary predicate operator v/val = valency +v /val = valency increase -v /val = valency reduction

Markers: GM = Goal Marker

Introduction

Treatises on Hebrew grammar are almost as numerous as on Latin. Ever since the early Middle Ages Hebrew has been studied by Jewish and Gentile scholars alike. For many generations Hebrew only served ritualistic purposes. Nevertheless, it was never as dead and static as commonly believed, though it is true that classic grammars are based on Biblical Hebrew. Hebrew was revived as a spoken language at the turn of this century. Contemporary Israeli Hebrew (which like its ancestor has been widely studied) differs from Biblical Hebrew in several significant ways. This study does not set out to add yet another description of Modern Hebrew to the already well-stocked shelves. Rather it aims at providing a point of contact between the language and the theoretical framework of this study: Functional Grammar. Functional Grammar, as developed in Dik (1978, 1980, 1987) and various later studies, is a linguistic theory in the so-called functional paradigm. The theories in this paradigm see language as a means of communication, and are interested in the structure and system of language in relation to its communicative function. This distinguishes them from theories within the formal paradigm, which are interested in the abstract system of language for its own sake. (This distinction comes from Dik 1983b.) In addition to this general aim, Functional Grammar strives for three standards of adequacy: pragmatic adequacy, psychological adequacy and typological adequacy. The demand for pragmatic adequacy results directly from its place within the functional paradigm, and requires that the rules governing verbal interaction should be incorporated in the theory. The demand for psychological adequacy implies that psychological models of linguistic competence and of linguistic behaviour must be taken into account. Finally, the demand for typological adequacy means that the theory should be able to give a good description of as many typologically varied languages as possible. (For more details on these standards of adequacy see Dik, 1978, 1983b, 1987.) This study has a double aim. On the one hand it sets out to test the typological adequacy of Functional Grammar (FG), and on the other to try and come up with new insights into the Hebrew language on the basis of the FG perspective. The first aim is clear enough since this is one of the few in depth studies of a Semitic language using FG as its theoretical framework. The study does not involve the whole FG model but only the lexicon and the predicate formation mechanism. More details on the theory of FG and the parts involved in this study will be given in chapter 2. The second aim, acquiring new insights into Hebrew grammar, is also limited to part of the language only, namely the verbal system. The verbal system was selected for two reasons: first, because it is a clearly defined topic occupying a central place in Hebrew grammar. Second, because it provides several topics which are interesting to consider from the FG point of view: the verbal patterns, in combining syntax, semantics and morphology, thus overlap conveniently with the predicate formation mechanism; as the representation of the roots is an inseparable part of this topic, FG will be faced with a type of problem that it has never been required to discuss before.

Throughout the study these two aspects of the study, the verbal system of Modern Hebrew (MB) and F6 are, intertwined. The innovative parts in chapter 3, which discusses the roots and the lexicon, centre on FG. Here some new theoretical suggestions are made on the basis of the problem of the representation of the roots and the organization of the lexicon. Chapter 3 is the contribution of MH to F6. Chapter 4 is a fairly straightforward application of the predicate formation mechanism of F6 to MH. At the same time, it tests the theoretical notion of 'productivity' on the basis of the processes in the verbal system. Chapter 5 is innovative with regard to the Hebrew verbal system, offering a different level of analysis from the usual one. Instead of regarding all the verbs as one whole system, they are divided into configurational sub-systems, to which the predicate formation mechanism is applied. In chapters 4 and 5 the term 'predicate scheme* is introduced for the first time in relation to Hebrew grammar. Chapter 5 is the contribution of FG to MH. The results, discussed in chapter 6, show that the combination has been fruitful. New insights into the verbal system of MH have been gained, and the application of the FG predicate-frames and predicate formation mechanism does help to form a fairly well ordered picture where many studies before saw only chaos. At the same time, FG also benefitted. Not only because the typological adequacy has been met, but also because some problem areas have been brought into focus.

Notes on the transcription

The Modern Hebrew alphabet has 27 consonants and 10 vowels, some of which are orthographic doublings; in earlier stages various consonants had a different pronunciation, but these are now lost. In spoken Israeli Hebrew there are differences in the pronunciation (of the consonants) according to ethnic origin. There is a dividing line between the pronunciation of the Ashkenazi (European and North-American) vs. Sephardi (North Africa and Middle East) groups in that the latter remains closer to the original Hebrew pronunciation. The transcription system used in the examples in this study follows the standard pronunciation (i.e heard on radio, taught in schools). The MH phonological system has 22 distinct consonant sounds, and 5 vowel sounds: Vowels:

MH has only short vowels, and no diphthongs; these are /a/, /e/, /i/ r /o/ and /u/, and a 'shwa' sound /g/. Consonants with the transciption used throughout the present study: /b/ - bilabial plosive /g/ - velar plosive /d/ - dental and alveolar plosive (voiced) /h/ - glottal fricative hi - labiodental fricative (voiced) (orthographically two different symbols) /z/ - alveolo-palatal fricative /h/ - pharyngal fricative /t/ - dental and alveolar plosive (voiceless) (two different orthographic symbols) /x/ - velar fricative /!/ - lateral glide /m/ - bilabial nasal /n/ - dental nasal /s/ - dental and alveolar fricative (two orthographic symbols) / ' / - glottal stop /p/ - bilabial plosive Itl - labiodental fricative Id - palatal plosive /k/ - velar plosive /r/ - uvular (or dental) rolled /δ/ - palato-alveolar fricative The following principles were adhered to throughout the study: 1. The transcription of words follows the standard pronunciation instead of the Hebrew orthographic signs. Thus no difference is made between the two orthographic signs for /t/, /v/ and /s/. 2. The transcription of roots follows the Hebrew spelling found in the sources (Barkali, 1980; Even-Shoshan, 1967). Thus a root may contain a vowel which is not distincly pronounced in the actual words derived from it. In such cases the vowel is transcribed only when the root is mentioned. For example the /h/ in the root r.a.h. 'see* is omitted in raiti 'saw Ip.s.' The transcription of a root consists of all its radicals , including those which are systematically omitted in the words derived from it. For

11

example the /u/ in the root s.u.t. 'sail', which is omitted in all the words derived from it, such as Sat 'sailed 3p.m.s.', ma So t Oar' or Sayit 'sailing 1 . 3. The transcription of the names of the binyaoim also follows the pronunciation and not the Hebrew spelling. Thus PAAL in fact consists of the three consonants /p/ / ' / /!/, which correspond to the radicals. An orthographic transcription would have been PA'AL, NIF'AL, PI'EL, PU'AL, HIF'IL, HUF'AL and HITPA'EL; the / ' / corresponds to the second radical. 4. Hebrew features a so called 'dageS', which changes or emphasizes a sound. Thus one of the /v/ symbols with dayeS is pronounced /b/, or the orthographic sign for /x/ with dayeS is pronounced /k/; the transcription here follows the pronunciation instead of the Hebrew spelling, as this seems to be more consistent. (Otherwise the Hebrew orthography would have to be followed throughout the transcription of each individual letter. The transcription of the pronunciation follows the tradition in other studies of MH, such as Berman (1978a) and Bolozky (1978)). 5. All roots in Hebrew consist of consonants only. Nevertheless, in some cases I transcribed one radical by a vowel. In this I mostly followed Barkali (1980), who added a vowel to the root in his list in order to avoid confusion with another root, which otherwise would seem homophonous but actually differs in one vowel. For example S.u.t. 'sail' and s.v.t. •strike 1 . Actually s.u.t. also consists of /δ/, /ν/, /t/, but this transcription would make it look identical to S.v.t. 'strike'. The transcription with an /u/ solves the problem and therefore seems to me worth some inconsistency. 6. The Hebrew alphabet does not distinguish between capital and small letters. Therefore, names in the Hebrew examples too are transcribed with small letters, and written with capitals only in the translations of the examples, or in quotations in the English text.

Chapter 1

Literature survey

1.0 Introduction As already stated in the introduction, one of the chief aims of this research is to test the typological adequacy of Functional Grammar on a Semitic language, in this case Modern Hebrew. By Modern Hebrew I mean the standard and colloquial contemporary Hebrew as used in Israel. (Note that much of what is said in this study about Hebrew applies with slight modifications to Arabic as well- hence the typological interest of this study. Such resemblances can be seen for example in Mitchell (1983) on Arabic.) Semitic languages differ in several points from the Indo-European ones to which FG has been applied so far. The one point relevant to this discussion is the morphological system. Semitic languages have a system whereby morphology, syntax and semantics are very strongly intertwined. The greater part of the vocabulary is formed by combining a consonantal root with a morpho-phonemic pattern. The root indicates a certain semantic field, and the patterns, the concrete form. The patterns are combined with the root and yield the concrete form (i.e. words). While nominal and adjectival patterns have only a morphological value, verbal patterns also carry syntactic values, expressing transitivity, causativity, reflexivity, etc. This is illustrated below, in (1) verbal and nominal forms derived from a root in Hebrew, in (2) a similar example for Arabic. verbal forms from the root k.t.b. 1 'write' (Hebrew): katav 'wrote' 3p.s.m. past tense ktov! 'write' imp. kotevet 'write' 3p.s.f. benoai (in MB present tense) yixtevu 'will write' 3p.pl.a. future tense nixtav 'was written* 3p.s.m. hixtiv 'made write, dictated' 3p.s.m. b nominal forms from the root k.t.b. 'write': katav 'correspondent, journalist' (m.) katava 'newspaper report* ( f . ) mixtav 'letter' (m.) ktovet 'address' ( f . ) maxteva 'desk' ( f . ) ktiv 'spelling' (m.) (2)a verbal forms derived from the root k.t.b. 'write' (Arabic): katab 'wrote' 3p.s.m. yiktib 'will write' 3p.s.m. 'ktib! 'write* imp. kaatib 'had written' 3p.s.m. b nominal forms from the root k.t.b. 'write' (Arabic): kaatib 'clerk' kitaab 'book' maktab Office, desk' maktaba 'library' maktuub 'written' (Da

13

In Hebrew grammars the verbal forms are called binyaaim, and the nominal forms are called aiSqalim; the terminology was introduced by medieval grammarians, like Ibn Ginah, Ibn Ezra and others. The verbal forms further conjugate for tense, number, gender and person. These will be presented later in this chapter. Thus a very large part of the vocabulary of Hebrew consists of such combinations of roots and affixes. This study deals with the verbal morphological patterns only, which in terms of FG covers the predicate formation mechanism and some expression rules. The expression rules realize the tense, number, gender and person inflections. They will not be dealt with in detail, only indicated. The verbal system, i.e. the patterns, their semantic-syntactic values and the relations between them are presented in the rest of this chapter in a literature survey on the biayania. This allows the reader to become acquainted with various views on the Hebrew verbal system, and puts the approach adopted in this study into a clearer perspective. At the same time, the repetition of many descriptions and examples will hopefully provide a strong enough informative background to the material discussed, so that the reader will not have to keep referring back. Much of the study will rely on the background information presented in this chapter. FG will be introduced throughout the remaining chapters, step by step as the subject requires, but almost all the information on Hebrew is concentrated in this chapter. I will first present the traditional view, then some contemporary views, and conclude with a presentation and justification of the view underlying this study. 1.1 The traditional vitw The traditional view on the binyaaia system is represented by Gesenius (1910) and Kimhi (13th cent.), and is still held by several ontemporary grammarians. Gesenius (1910), Driver (1879) and Kimhi in fact did not describe Modern Hebrew, but Biblical and post-Biblical Hebrew. Nevertheless, this view is presented here not only for the sake of comprehensiveness. Some contemporary grammarians like Sasson (1976) still adhere to it, and moreover, it still provides the basis for many Modern Hebrew school books. The traditional view, in other words, remains widely accepted. For the sake of clarity, the FG terminology is used throughout, even when presenting descriptions based on other grammars. Their widely varying terminology will be quoted in brackets. 1.1.1 Geseniu· Gesenius (1910) distinguishes between the following three groups of verbal forms: 1. the verb in the PAAL pattern (called 'verbal stem', 'primitive verbs'), where we find only the root consonants without any addition. For example (3)

raca 'wanted 1 from the root r.c.h. + the vowels /a/

2. the other verbal forms which are derived from the PAAL ('verbal derivates 1 ) either by change of vowel(s) and/or addition of consonants. For example

14 (4)

raca

> > >

nirca ( KIFAL -B2) rica (PIEL· -B3) hitraca (HITPAEL· -B7)

It aay be of interest to note that PAAL comes from the root p.'.l. 'action*. The names of all the binyaaia are in fact derived from this root. 3. denominatives: verbs2 ('primitive* or 'derived') which are derived from nouns or even from particles. For example (5)

«ore« 'root* —> seres (PIEL -B3) 'make rooted' 1 —> hisris (HIFIL -B6) 'make rooted

One of the forms, the PAAL (also called Qal) is regarded as the ground form, and the other forms are derived from it. The derivation can consist of a. change of vowel: PAAL—> PIEL (CaCaC —> CiCeC ) b. addition of prefix: PAAL·—> NIFAL (CaCaC —> NICCaC) PAAL· —> HIFIL· (CaCaC —> HICCiC) PAAL· —> HITPAEL· (CaCaC —> HITCaCeC) c. strengthening the middle consonant: PAAL —> PIEL·; the verbs in the PIEL· pattern have a marker which emphasises the second root consonant (radical), called dageS 'emphasis marker'. d. repetition of one or two of the stem consonants: CaCCaC, CiCCaC (we do not count these as separate biayania, see section 1.5.2). The whole binyaaim system is represented as follows: (illustrated by the root q.t.l. 'slay') (6)

Active 1. Qal/PAAL qatal 'to kill' 2. NIFAL niqtal 'to kill oneself 3. PIEL· qitel 'to kill many, to massacre* 5. HIFIL hiqtil 'to cause to kill' 7. HITPAEL

hitqatel 'to kill oneself

Passive (rarely passive) 4. PUAL qutal 6. HOFAL /HUFAL3

huqtal (very rare HOTPA'AL·)

Gesenius notes that very few roots occur in all these seven binyanim. also gives an -in his words - more satisfactory division, PAAL: the basic pattern The intensive: PIEL, PUAL and HITPAEL Th· causative: HIFIL, HOFAL (and SAFEL/SIFEL which will be discussed in section 1.5.2) 4. The reflexive or passive: NIFAL. Gesenius (ibid: 101) says of the combination root + pattern that "the roots are mere abstractions of stems in actual use and are themselves not used. They represent rather the hidden germs (semina) of the stems which appear in the language." He 1. 2. 3.

As for the biayanim, he gives the following characterizations: PAAL (QAL·) (Bl): basic NIFAL (B2): a. resembles Greek middle voice: nistar 'is hidden', niSmar 'is guarded'; expresses emotions which react upon one's mind or which one passively accepts: nibau 'is reconciled', neenah 'sighs'

15

b. expresses reciprocal or mutual action: nidberu 'agreed with each other', ailbeau 'fought each other', oo'ac 'consulted someone' c. active: neenab 'sighed', ailbaa 'fought* d. passive of QAL: nixtav 'was written', niSal 'was asked* e. passive of PIEL or HIFIL (if there is no PAAL): aixbad 'was offered* PIEL· (B3) and PUAL (B4): a. intensification, strengthening and repetition of the action: hilex 'walked- iterative', Siber 'smashed' ('broke -intensive'); b. causative: gidel 'made grow', cidek 'caused to be right', kicec 'made shorter'; c. PUAL: passive and participle of PIEL. kucac 'was made shorter', gudal 'was made to grow'; HIFIL (B5) and HUFAL (B6): a. causative of PAAL: birSi'a 'made guilty', hod 'caused to get out; b. transitive of intransitive PAAL: c. HIFIL occurs in stems expressing inchoative: higbiha 'made tall/ made long', bifriab 'caused to blossom'; d. express incipience of a certain condition and its continuation: beemin 'believed', biSkit 'caused to be quiet', birgi'a 'caused to be calm'; e. action in some particular mental direction: behti 'caused to sin', beytiv 'caused to be good, improved', biskil 'made clever'; f . denominatives expressing the drawing out, the production of a thing blirii 'caused to be rooted', biSmin 'caused to be f a t ' . 4 The HUFAL expresses primarily the passive of HIFIL: buSraS 'was made to become rooted'; sometimes it is also the passive equivalent of the PAAL: nakam 'avenged* - bukau 'was avenged' HITPAEL (B7):° a. primarily reflexive of PIEL: bitnakem 'revenged himself, bit'aSer 'enriched himself b. equivalent of PAAL: avel - hitabel 'mourned' c. reciprocal: bitrau 'saw each other', bityadedu 'got befriended with each other' d. middle ('action less directly affecting the subject and describes it as performed with regard to or for oneself * ) . According to Gesenius in such cases the HITPAEL takes the accusative: bitnacel 'apologized', bitparek 'got dismantled', bitpalel 'prayed', bictayed 'supplied himself; in Ködern Hebrew, however, only bitpalel 'prayed' can occur with a Goal. e. passive (but quite rare): biStakeab 'is forgotten' f. inchoative ('denominative with reflexive meaning'): bityabed 'became Jewish' In addition to the derivation/conjugation of the binyaoia, Gesenius presents the following paradigm of inflection for gender, number and tense. (Gesenius, who described Biblical Hebrew, mentions aspect: perfect and imperfect instead of tense.) (7)

Perfect 3m. 3f. 2m. 2f. Im/f.

singular c.c.c. c.c.c.-ah c.c.c.- ta c.c.c.-t c.c.c.-ti

plural 3m/f. c.c.c.-u 2m. c.c.c.-tern 2f. c.c.c.-ten Im/f. c.c.c.-nu

16

(8)

Imperfect 3m. 3£. 2m. 2f. Im/f

singular i-c.c.c t-c.c.c t-c.c.c. t-c.c.c. a-c.c.c.

plural 3m. 3f. 2m. 2f. Im/f.

i-c.c.c.-u t-c.c.c.-na t-c.c.c.-u t-c.c.c.-na n-c.c.c.

There is one other for·, the so called benoni. In Biblical Hebrew this was not really a verbal form, but rather something between participium and nominal. This is clearly illustrated by the following example: (9)

...parot habasan a£er behar somron ha'oSkot dalim kine of-the-Bashan that in-mountain Somron that-oppress poor harocecot evyonim haomrot leadoneyhem... (Amos, 4:1) that-crush poor that-say to-their-masters... 'Hear this word ye kine of Bashan, that are in the mountain of Samaria, which oppress the poor, which crush the needy, which say to their masters...' (King James translation)

In Modern Hebrew, however, it is increasingly used for denoting present tense. (In fact the use of the benoni as praesens results from a historical process which started already in the Mishnaic period, as described in Gordon (1982)). The benoni differs from the perfect and imperfect. One difference is that it can be made negative with the particle eyn 'no', used otherwise only with nominal or adjectival forms, never with perfect or imperfect. Its inflectional paradigm too resembles the nominal one, in that benoni verbs inflect only for number and gender, not for person. The verbal system has three other forms: the imperative (only in the active patterns) the so-called verbal noun which resembles the gerund, and the infinitive: the stem prefixed with li. For example (10)

lex! 'go' (imperative) lixtov 'to write' (infinitive) halixa 'going, walking' (verbal noun)

The inflectional paradigms for number, gender and person are the same in Modern Hebrew: the perfect is used as past tense, and the imperfect as future tense; the present tense is expressed by means of the benoniwhich is in BH an intermediate form, partly verbal and partly nominal. To sum up, Gesenius describes the verbal system of Biblical Hebrew as follows: 1. it is based on binyanim which are systematically related to each other in expressing syntactic and semantic features; 2. one of the binyanim, the PAAL, is basic, both morphologically and syntactically-semantically; all the others are derived from this pattern; 3. there is a distinction between derivation (in Gesenius's terms 'conjunction') and inflection; the former is the system underlying the binyania, the latter marks number, gender and aspect. Driver (1879) deals primarily with the tenses in Biblical Hebrew, but the verbal system underlying his study is the same as in Gesenius (ibid). 1.1.2 Klmhi Kimhi wrote his grammar of Hebrew in the 13th century, describing Biblical Hebrew. His main contribution was in the field of phonetics,

17

consisting of a new and remarkably accurate description of vowel changes. His description of the biayaaim system is based on eight biayauia. Kimhi (ibid) believed that all the verbs were formed from three consonantal roots. The further description is as follows: PAAL (QAL) in the third person masculine perfect (past tense in Mfl) is considered to be the basic form from which all the rest are derived. NIFAL: the passive or reflexive of the QAL·, and if the QAL is intransitive it can also be the passive of FIEL or HIFIL (as in bikriv 'sacrificed 1 - nikrav 'was sacrificed'} PIEL·: its distinctive character is an emphasis (dageS) in the second radical. It is transitive, causative or expresses the intensified meaning of PAAL or HIFIL·. PUAL: exclusively the passive of FIEL. HIFIL: causative HUFAL: passive

of HIFIL·

POEL: transitive or causative, like the FIEL. (eg. yoda'ati 'I knew* (IS, 12:3), veyolada 'and begoth' 3p.m.s. (Ju, 13:5) ) (Note that Gesenius does not see the POEL as a separate pattern.) HITPAEL: reflexive of POEL; sometimes expresses the idea of pretending to do or be what is expressed in the predication, (eg. bitabel 'pretended to be ill' 2S, 13:5)

1.2 Th· 'lexicalist' view The descriptions presented so far featured the biayaaim as the organizing principle of the Hebrew verbal system. That means, a concrete verb is seen as follows: Both components are given equal status, and the morphological system of prefixes and suffixes marking gender, number and tense is seen as secondary. This section will feature a different view. The term 'lexicalist' was coined specially for this presentation (it is not used in the literature). For this theory attributes the same functions to the lexicon that the other theories attribute to the biayaaia system. The leading representatives of the lexicalist approach are Oman (1969, 1971a, 1971b, 1971c, 1978, 1979) and to a lesser extent Schwarzwald (1974, 1975, 1981a, 1981b). 1.2.1 Oman Oman proposes a unified morphological system, which would give a similar account of the miSqalim (the nominal patterns) and the biayaaim system. He suggests that every word which entered Hebrew as a complete word (for example Sulbaa from Aramaic, for which no root can be traced) has the following morphological structure: (11)

word * stem + formative

The formatives are all added to the stems by means of morphological rules. A stem is not a root; it is identified by means of phonological criteria as that part of the verb or noun which remained unchanged throughout the whole derivational and inflectional paradigm. There are two types of formatives: those which are entirely regular and predictable, remaining constant throughout the whole verbal system; and those which are irregular and by and large unpredictable. The former are the inflections, as presented in tables (7) - (8) above; the latter are

18

the biayanim (for the verbs), as listed in (6). These formatives apply only to a rather small percentage of all verbs: according to Schwarzwald (1981b) to only 2.3%. In addition, the PAAL, NIFAL, HIFIL and HITPAEL· nay have several functions. It is impossible to predict for a certain root which of the possible functions a given pattern will have. (For example, the HIFIL pattern can express transitivity and causativity; there is no general rule which could predict whether a given root will express the one or the other.) This unpredictability means, according to Schwarzwald and Oman, that the system is not regular. At least, not regular in comparison with the system of inflection for number, gender and tense (which is always totally predictable; the suffix au means in all binyaaim 1st person plural past tense). Oman bases his distinction between the two types of formatives on this comparison, and concludes that the biayaaim should not be seen as a systematic principle. Rather, the syntactic and semantic information coded in them (like passivity, reflexivity, reciprocity etc.) should be simply listed in the lexicon. The system which Oman suggests is purely morpho-phonologically based and motivated. The verbal system has three sections: past, present (beooDi) and future. The stem is identified by working 'backwards'as it were, removing all the prefixes and suffixes. What remains morphologically and phonologically unchanged is the stem of a certain verb for the section in question. The same root may have a different stem for the past, the beooni, and the future.· For more details, the reader is referred to Oman (1971a, 1971b, 1971c, 1978, 1979). 1.2.2 Schwariwald Schwarzwald's standpoint is less clear. She supports Oman in principle,, and even reports an experiment (Schwarzwald, 1981c) showing that average speakers of Hebrew have difficulty recognizing the root, in particular of 'weak verbs', 7 and that the biayanim system is in fact not regular and cannot be the basis of predictions. Consequently, the semantic-syntactic information which the traditional view attributed to the binyanim should be incorporated in the lexicon. The binyanim then are merely morpho- phonological formatives, like the inflections. Nevertheless, in Schwarzwald (1975) the NIFAL, PUAL and HUFAL are seen as passives of PAAL, PIEL and HIFIL·. This, however, is a regular relation, represented as a grammatical rule. (Note that it applies according to Schwarzwald, only to those cases where the NIFAL is the 'passive' of the PAAL, and this is by no means always the case.) In addition, the lexicon of ME suggested in Schwarzwald (1975) regards the binyanim as capable of expressing syntactic features. The main advantage of the lexicalist approach is that it presents a unified and simple system for the verbal and nominal morphology. No doubt, it would be simpler to base school books on this approach, as Oman and Schwarzwald suggest. However, in terms of grammatical description, such simplicity would amount to throwing out the baby with the bathwater. There is evidence that the speakers of Hebrew see the binyanim as something more than just morphological formatives, like the inflections for number, gender and tense. Such evidence is presented by Waiden (1982) and Berman (1981, 1982b, 1983) for child language. ( For details see section 1.5.1). Furthermore, the new verbs which are added to the system, if transitive are in PIEL· or HIFIL·, and if intransitive, in HITPAEL·. (see Bolozky, 1978). If these patterns did not have any syntactic value for the speakers, then we would also get new transitive verbs in the NIFAL· or the HITPAEL, for example.

19 Finally, the use of various verbal forms is bound to certain grammatical contexts which can be described systematically by using the notion 'valency 1 . This notion, in turn, is based on the existence of the biayaoia as a systematic way of expressing syntactic and semantic features of the verb. (Cf. chapters 3, 4 and 5).

1.3 Traditional and Transformational Generative view· This section discusses yet another way of dealing with the verbal system: a Transformational Generative approach advocated in Berman (1975a, 1975b, 1978a, 1978b), a logically based approach presented in Ariel (1972), and the traditional approach based on Gesenius, advocated by Sasson (1976), Blau (1967, 1972), and Goshen-Gottstein et al. (1967). In principle they all share the view that the binyaaim concept should be retained as the organising principle of the verbal system in MB, unlike the lexicalist view presented in the previous section. 1.3.1 Berman Berman's (1981, 1983) evidence is derived from child language and shows that children learning Hebrew as their mother tongue are aware of the syntactic value of the bioyaaia. It turns out that the mistakes they make never cross the so called 'transitivity line': children may use one transitive pattern when they should use another (say HIFIL instead of PAAL·), but they will never wrongly substitute a transitive pattern for an intransitive one (say PIEL instead of NIFAL). In other words, the transitive patterns PAAL, PIEL and HIFIL are already separated at the age of 45 from the NIFAL, PUAL, HUFAL and HITPAEL. Berman (1975a) lists the following seven biayaaia as forming the system (the distinction between 'main' and 'secondary' function is based on frequency of occurrence and productivity; the 'main 1 functions are productive, and the 'secondary' ones are 'idiosyncratic'. The secondary functions are listed in descending order of frequency): (12)

biayaa PAAL NIFAL·

main function basic (+/-tr) 'passive' of PAAL

PIEL·

basic (+tr)

PUAL HIFIL

'passive' of PIEL· causative of PAAL

HUFAL HITPAEL·

'passive' of HIFIL (-tr), middle of PIEL·

secondary functions middle of HIFIL· basic (intransitive) reciprocal +tr or specialized to PAAL inchoative +tr of verbs that are basic in NIFAL reflexive reciprocal inchoative iterative basic

The following sentences (from Berman (1978b)) illustrate the syntactic processes reflected in the biayaaim forms of verbs with a shared root:

20

(i) voice active: (13)a dan imen (B3) et hacevet. Dan trained GM the-tearn 'Dan trained the team', passive: (13)b hacevet uman (B4) 'al yedei dan. the-team was-trained by Dan 'The team was trained by Dan' middle: (13)c hacevet hitamen 'took' Past take V (xl)Ag (x2)Go Subj —> 'taken'

the verb 'take' forms, such as realization of the arguments

39

For nominal predicates too, the basic predicate is the stem. For example (14)

book N (xl)0

Such a nominal predicate may be used to form a term, which can then be inserted into an argument slot in a predicate frame. Therefore, in its final realization the term operator(s) too decide on its final form, which final form can for example be the plural 'books'. In languages with a case and gender system the final form of nouns (and adjectives) is also influenced by their semantic function in the predicate frame, λ clear example is Latin, where we never encounter the stem form of neither verbs nor nouns, but only inflected forms. The inflection, although obligatory, is a separate component as it were. Thus for these languages we have two levels of abstractions with regard to the predicates: the somewhat abstract level of stems (which in some languages can overlap also with a concrete manifestation of the stem) and the concrete level of the actual word as we encounter it in use. In Hebrew, however, there is an additional level of abstraction: the root, λ root, such as k.t.b. 'write 1 , is more abstract than a stem. It stands for a semantic field, that of writing for example. From it both verbal and nominal forms can be derived. At the same time, Hebrew has no forms which are strictly equivalent to stems since every form except for the root includes not only the morphological pattern, but also number, gender and tense. Comparing Hebrew to English we get the following picture:

(15)

English

abstract form stem

'write'

concrete form (=word)

Hebrew

FG representation English Hebrew C.C.C.

root: k.t.b.

write

write V (xl) (x2) katav

Pres write V (xl)AgSubj (x2)Go

Pres k.t.b.V-Bl

(xl)AgSubj (x2)Go

Note that in Hebrew the concrete form is obligatorily specified also for number and gender. However, according to the FG model the specification of these features occurs only in the very last stage, by the expression rules. Therefore, the problem arises what is the form of the predicate? It cannot be a concrete word, for it includes information which is inserted only at a much later phase. Nor do we have a stem in the strict sense of English. Thus the term predicate-frame is used in this study to refer to entities like in (16) below, but these are not striclty stems.

(16)

katav V-B1 (xl)Ag (x2)Go

The predicate-frame for MH includes the verb, here in the 3rd person singular masculine, morphologically the simplest form in that it is the closest to the root (where for simplicity's sake I ignore here that it is also specified for number and gender), the biayaa and the argument slots with their semantic functions. As we shall see later throughout the text, this is necessary for a clear exposition of Hebrew for non-speakers of the language.

40

There are two more forms which the reader will encounter, and which are new within FG. These are (17) (18)

k.t.b. V-B1 (xl)Ag (x2)Go C.C.C. V-B1 (xl)Ag (x2)Go

(17) is a so called 'predicate-scheme', and (18) is a 'generalized predicate scheme*. A predicate-scheme is a predicate-frame, only not for a concrete verb («word), but for a root + binyan. That is, it contains the same information as the predicate-frame. But instead of being a concrete combination of root + pattern, it provides an abstract representation of the predicate-frame that a root would have in a given binyan. A lexical entry consisting of a predicate-scheme can consist of several predicate-frames, each of a different root. A generalized predicate-scheme is simply used in talking about properties of predicate-schemes in general, without mentioning a particular root. This enables focusing on the properties of specific binyanim, as illustrated below: (19)

PAAL transitive: C.C.C. V-B1 (xl)Ag/Fo (x2)Go/Compl PAAL intransitive: C.C.C. V-B1 (xl)Pos/Proc [ (x2) LOG, Temp, etc] HIFIL causative: C.C.C. V-B6 (xO)causer (xl)causee [ (x2)Go ] ( [ ] = optional argument)

Note that in so far as content and function in the description of the Hebrew verbal system are concerned, the 'predicate-scheme' is similar to a 'predicate-frame'. The only difference is that it is more general, and applies to a root + binyan instead of to a specific word. This suggests that the notion of predicate-frames is applicable to NH, but technically it is too narrow to describe Hebrew, or other Semitic languages which have roots and patterns. Hence the concept of 'predicate-scheme' had to be introduced, which is more general than the predicate-frame. In a lexicon of Modern Hebrew (or any other Semitic language) both concepts are necessary. The predicate-schemes will serve to illustrate the valency relations in the biayaaia system. Thus, in contrast to such languages as English or Dutch, predicate formation rule-like operations will be applied not only to actual predicate-frames, but also to the more abstract level of predicate-schemes. This leads to the view on which this study is based, namely: The organizing principle of the MH verbal system is based on the notion of binyanim and of predicate-frames. The binyanim can be represented in predicate frame-like constructions: predicate-schemes. Hence, the operations applying to predicate-frames are assumed to be applicable also to the biayaaia system. This assumption leads to one of the central hypotheses, namely: Since some of the processes involving the derivation of predicates can be expressed in terms of valency change, some of the derivational processes concerning the binyanim can be expressed in terms of valency change too. Hence, the derivational relations between the binyanim can be described in terms of changes of valency. In addition this applies not only to the concrete verbs derived from the same root, but also to the binyanim as a system. This study elaborates the view that the binyanim system is regular and systematic (in contrast to the lexicalist view described in chapter 1 section 1.2). However, the systematism in the binyanim system should not

41

be searched on the top level of all the verbs, but on a lower level of the so called 'configurational sub-systems' (described in detail in chapter 5, section 5.2.).

2.6 Summary This chapter presented a short background on those parts of Functional Grammar which are relevant to this study. Namely, the lexicon, the predicates and predicate formation. They have been treated here only in general. Specific issues connected to these topics, such as for example the question of productivity connected to the predicate formation rules, are treated in detail only at that point of the study where they become relevant and necessary. The topic of predicate schemes and generalized predicate schemes too, has been touched upon only by way of introduction. These terms, however, will become clearer simply by following their use throughout the study.

Chapter 3

The lexicon and the roots 3.0 Introduction The previous chapters presented the Hebrew morphological mechanism whereby roots are combined with patterns, in the formation of both verbs and nouns/adjectives. This research deals primarily with the verbal system, its syntactic aspects and derivational characteristics. But no serious study of Hebrew can avoid discussing the status of the roots. Furthermore, incorporating the roots in this study has consequences for the theoretical framework: Functional Grammar. Figure (1) in chapter 2 shows clearly that the lexicon is the most logical place for the roots. Therefore, the roots will be dealt with in this chapter together with the representation of the bioyanim in the lexicon. The lexicon in FG as well as some general principles of the representation of the Hebrew verbal system in it were discussed in chapter 2. This chapter goes in further detail into the topics of the lexicon for MB, its components (predicate-frames and predicate-schemes), its organization and the problem of the representation of the roots. 3.1 'Dictionary' vs. 'lexicon' The term 'dictionary' is usually used for denoting the list of all the words existing in a language. It may either consist exclusively of words current at the time of compilation, or include words which have become obsolete, λ dictionary is organized alphabetically. The aim of a dictionary is to supply (a) information about all the words which exist in the language, and (b) the meaning (and use) of these words. The term 'lexicon* on the other hand, is usually used to denote a part of a linguistic theoretical model of a given language. As such it is closely linked to the 'grammar* (or syntax) part of the model, i.e. to the set of syntactic, morphological and phonological rules as formulated by the theory. Therefore, whereas a dictionary has an independent existence, a lexicon does not. Although the lexicon, like the dictionary, may consist of the list of words existing in a language, its function is different. It is primarily part of a theoretical model. Hence it is perfectly plausible to arrange a lexicon non-alphabetically. Within the framework of a given theoretical model a lexical entry might demand the addition of certain information which is not included in a dictionary. Of course, compilers of dictionaries also often hold certain linguistic theoretical views, which may implicitly or explicitly influence their dictionaries. This is usually evident in the introductory summary where the grammar of the language in question is presented, and can also occur in the explanation of the use of the words. A survey of the approaches to Hebrew will show that the linguistic theoretical view of the compiler of a Hebrew dictionary plays a much more prominent place than that of a compiler of say an English or French dictionary. In addition to the general theoretical distinction, there is a special reason for distinguishing between 'lexicon' and 'dictionary' for Hebrew. In Indo-European languages, like English or Dutch, corresponding entries in the lexicon and a dictionary both consist of the basic predicate (which is a stem) in its frame. In a lexicon for HH, however, this is not

43

the case. For Hebrew a lexical entry does not have to consist of concrete words, and cannot consist of lemmata, as these do not exist (in the same sense as for English). Therefore, the lexicon suggested in this study is very clearly part of a linguistic theoretical model and should not be confused with a dictionary. This approach is also new for F6, which has not yet been applied to languages with a morphological system similar to Hebrew (at least not in treating the lexicon). The lexicon suggested in Dik (1980) can be seen as an extended dictionary; the lexicon for MH cannot, because its entries do not consist of lemmata. So the distinction between dictionary and lexicon is an innovation both for the language described and the theory of FG. 1 I wish to stress that this study (sections 3.4 and 3.5) proposes a lexicon and not a dictionary. 3.2 Approaches to the Hebrew lexicon There are four approaches to the Hebrew lexicon: (i) all the words are derived from roots (ii) most of the words are derived from roots (iii) only the verbs are derived from roots (iv) every word is an independent entry. Approach (i) can be found only in old studies compiled in the llth century, like 'The book of roots' by Rabbi Jonah Ibn Ginah. Such studies are based on Biblical Hebrew and include under each root all the words (verbs and nouns) derived from it. There are no loose words as such. The 'Book of Roots' is part of a two volume work. The first volume is called Kitab al-Luma ('book of coloured flower beds') and is an exhaustive grammar of Biblical Hebrew; the second volume, Kitab al-Usul ('book of roots') is a complete list of the roots in the Bxble, followed by illustrations and explanation. In the case of obscure roots, comparison is made with Arabic. The morphological patterns do not occur at all as separate items. The Book of Roots is an alphabetically ordered list of all the roots and words formed from them. The list of words is not exhaustive; nor are they accompanied by any grammatical information, like category or gender, which we do find in all the modern dictionaries. This work, however, is a dictionary, though not in the current sense of the word. Approach (ii) is also to be found only in a few old dictionaries (and Bible concordances), such as that of F. Brown, S.R. Driver & C.A. Biggs (1907). (Quoted from Schwarzwald 1975.) Here all the words, verbs and nouns whose roots are known are listed in the entries of the roots; there is a small group of words whose root is unknown. These are listed in separate entries, as words. Here the roots are represented independently as entries, and the binyanim and miSqalim are only implied. Most of the dictionaries of Modern Hebrew are based on approach (iii). This means that there are two types of entries: roots, represented independently, followed by all the verbal forms derived from them (appearing as words; the binyaoim and miSqalim are not represented independently, nor is the bioyan of a certain form stated). All the nominal, adjectival and other non-verbal forms occur as words. Thus the non-verbal forms appear as independent words, in alphabetical order, whereas all the verbs appear alphabetically according to their first root consonant. The result is inconsistent, because in beexil 'feed' the [he] is treated as the HIFIL prefix, and beexil occurs under the letter /a/, due to its root: a.x.l. At the same time, aaaxal 'food', which is derived from the same root, occurs under the letter

44

/m/; here [ma], the prefix of the aiSqal {= nominal pattern) maCCaC is treated as part of the word. (CCC stands for the root consonants). The fourth approach, namely listing all the forms as separate words is adopted in one dictionary only, Sivan & Levenston (1967). (Quoted from Schwarzwald 1975.) Here neither the roots nor the binyaoim get an independent representation. This approach is also shared by Oman (1971a, 1971b), whose view on the MB verbal system was presented in chapter 1. In Oman's view a dictionary of Modern Hebrew should consist of word entries as in Indo European languages. Schwarzwald in her survey of these approaches (1975) uses the Hebrew word uilon, which can mean both 'dictionary* and 'lexicon*. If the distinction presented above between dictionary and lexicon is applied, then approaches (i) (ii) and (iii) present a lexicon, whereas approach (iv) presents a dictionary. The most widespread approach is (iii), which is also the basis for most grammars and dictionaries of HH. Here the formula root + pattern which holds both for verbal and for nominal constituents, is used only for verbs. The reasons which are usually given are: 1. The group of verbal patterns is small (only 7-9); it forms a closed group, its meaning is relatively easy to define and there is a relatively high consistency in the verbal system. 2. The group of nominal patterns is large (some 120 different patterns); it forms an open group (i.e. new miSqalim are still being formed); and it is highly inconsistent and unpredictable both in occurrence and in the meaning with the various concrete roots. There is only a very small group of miSq&lim with a somewhat consistent meaning, such as maCCeCa 'a place for doing CCC', or maCCeC 'a tool for doing CCC'. Schwarzwald (1975, 1980) suggests that verbal and nominal forms should get the same representation: root + pattern, followed by phonological and syntactic features, and a meaning definition. 2 The verbal forms are listed in the lexicon only if they are idiosyncratic, otherwise there is a rule by which they are derived from the basic pattern (which is the PAAL). There are two types of rules: 1. Those that represent processes holding for the whole verbal system. These are 'grammatical rules'; they derive only the NIFAL, PUAL and HUFAL as the passive of the active transitive PAAL, FIEL and HIFIL respectively. The verbs in these binyaoim are, therefore, not listed separately. 2. The so called 'minor redundancy rules' for deriving the other biayaoim, whose function varies depending on the root. (I.e. HITPAEL is reciprocal of PAAL for root X ) . The root in this pattern is listed separately, as a word. A lexical entry includes the following information: a. root + basic binyao b. the list of patterns in which the root in question occurs, with syntactic information (transitivity), meaning and phonological changes in the root consonants (when there are any); c. the 'minor redundancy rules' applying to the root in question, and the actual verbal forms resulting from the application of these rules, i.e. the idiosyncratic occurrences of the root. 3 This approach, at least as far as the representation of the binyanim is concerned, is in principle very similar to Herman's (1975). Berman (1975) clearly presents a lexicon based on a linguistic theoretical concept, and not a dictionary. Herman's analysis is based on the Chomsky (1965) version of the theoretical framework of Transformational Generative Grammar. Berman (ibid) suggests representing the verbs in the lexicon using the binyania system. An entry will start

45

with a root in the PAAL, which is seen as the 'basic pattern'; all the other patterns are derived from it. This 'root in the PAAL' is accompanied by the syntactic environment in which it occurs, namely the noun phrases (NP) which it takes, and the semantic features of these HP's. The other binyania featuring the root in question, are derived from it. If they have the general productive syntactic value (i.e. the syntactic value considered typical for that pattern), then the derivation is carried out by means of redundancy rules, based on Chomsky (1965). Otherwise, the verbal forms in question are included in the lexical entry. 4 ( C f . chapter 1 section 1.3, table (12).) Berman's approach is similar to Schwarzwald's in the following points: 1. Both suggest distinguishing between the productive, regular syntactic values of roots in certain binyanim and the unproductive, idiosyncratic values of those roots in other binyanim. The former category is to be taken care of by rules that are part of the grammar, not the lexicon. The latter category is to be listed in the lexicon with the appropriate 'redundancy rules' (in Berman's terms) or 'minor rules' (in Schwarzwald's terns). 2. Both Schwarzwald and Berman try to account for all the roots as one whole system. That is, to give rules and descriptions which have to hold good for the whole verbal system. This, however, is practically impossible. Therefore, in attempting to overcome this difficulty Schwarzwald and Berman have to either limit their description to the main functions of the binyanim only (as Berman does), or to renounce the possibility of formulating more than very limited rules (Schwarzwald). An FG lexicon of a Semitic language would have the following components: a. the roots; b. the verbal and nominal predicates (in predicate frames). The next two sections describe these components. 3.3 The root

When looking at the lexicon of a Semitic language, in this case Modern Hebrew, the first question is how to deal with the root. In answering this question, it must be decided whether: 1. the root should be represented as an undivided entity, individually in the lexicon? 2. if so, how? Although the roots are a clearly distinct morphological unit in Hebrew, it does not follow from this that they have to be represented as such. One of the possibilities, therefore, is that lexical entries should consist of predicate-frames, where the root is considered as a morphological and semantic component of the predicate but is not represented separately. According to the lexicalist approach all words derived from roots are represented in the lexicon as separate entries. In these cases the roots are not separately represented; they are only one of the components of the derivational rules. This approach, however, has been rejected. (See chapter 1 section 1.5.1 .) As the aim of this study is to test whether the FG model is (directly) applicable to the MH verbal system, the Hebrew lexicon, like that of other languages described by FG, should consist of entries which are basic predicates. Basic predicates in their turn, are all lemmata (which may overlap with concrete words of the language), but are in any case not abstract entities that are never encountered (be it only as part of words) in usage. Since Hebrew has no lemmata, a possible

46

choice (as indicated in chapter 2) would be of concrete words. However, choosing actual words as the entries of the lexicon would create the following problems: 1. Which criteria should be used in deciding whether the basic predicate of a certain root is verbal or nominal? 2. Any actual word in Hebrew is automatically inflected for person, gender, number, and for verbs also tense. Hence the basic predicate would include features that do not belong there, but are introduced only much later by the expression rules. There is a convention by which the past tense 3rd person masculine singular is commonly used as a 'representative' form, since it is morphologically the simplest. The representation of the verbs in 3rd. person singular masculine in this study is also based on this convention. Ad 1 Possible criteria for selecting a nominal or verbal predicate as basic are the following: (i) Quantitative criterion, i.e. whether a root occurs more frequently in verbal or nominal patterns. The problem here is that many roots occur in more or less the same number of bioyanim and miSqalim. (ii) Morphological simplicity criterion, i.e. whether the morphologically simplest form of the root is verbal or nominal. This also meets with several drawbacks: a. often there is no difference between morphological complexity between the verbal and nominal form, as in

(1)

Sefex N 'water spring*

-

safax V 'poured 1

b. applying this criterion leads to inconsistencies as demonstrated below; in some cases (2a) the verbal, in others the nominal is simpler (2b): (2)a b

Saal V Samar V kanas V katav V

'asked' 'guarded' 'fined' 'wrote'

- seela N 'question' - miSmeret N 'a guarding period' - knas N 'a fine' - ktav N 'writing'

(iii) Semantic simplicity criterion. If it is easier to paraphrase a verb in nominal terms derived from the same root than a nominal form in verbal terms, then the nominal predicate is basic (or vice versa). This criterion is sometimes effective, as in (3)

even N 'stone* zer'a N 'seed*

-

iben V 'turn into stone' zar'a V 'sew, plant seeds'

but it is also often not effective, as in (4)

perah N 'a flower* Sever N 'a break' remez N 'a hint'

- parah V 'flowered, blossomed1 - Savar V 'broke* - ramaz V 'hinted'

In the last three examples in fact none of the three criteria can be applied effectively: the nouns and verbs are morphologically and semantically equally basic and there is no clear quantitative difference between the number of verbal vs. nominal patterns in which they occur, (iv) Historical criterion: in some cases the etymology of a word indicates cleary enough whether the verbal or nominal form existed first. This criterion is applicable mostly to innovations, because there the

47

origin can be traced clearly. Innovations in Modern Hebrew mostly concern the formation of denominal verbs, a process which is very productive in Modern Hebrew. Ad 2 As mentioned earlier, every verb in Hebrew is inflected for number, gender, person and tense. Therefore, the basic predicate Samar 'guarded'

(5)

Samar V-B1 (xl)Ag (x2)Go

'guard'

includes more information than 'guard' does in English; it also includes '3rd. person masculine past tense'. And due to the fact that even 'stone' is singular masculine, the predicate frame (6)

even N (xl)0

also includes the information 'masculine singular*. This is information which does not belong in the basic predicate-frame, but is only introduced when the expression rules are applied. But it is impossible to omit these features in an actual word in Hebrew, unless that word is redefined as a lemma. This, however, would lead to another methodological problem, namely when is a word a word and when is it a lemma? In Hungarian for example the lemma /kapcs/ 'connect' of the verb kapcsoloi 'to connect 1 , or the noun kapcsolat 'connection', differs formally from any concrete word occurring in the language, and is thus suitable as a lexical entry. There is no such lemma in Hebrew. Anything which is more than a root or an abstraction represented as root + pattern (eg. h.l.x. + PAAL 'walk' + PAAL) is a concrete word. In practice, using lemmas as lexical entries would simply involve redefining a word to be a lemma, so as to redefine it later again as a word. Furthermore, it is undesirable to include these features in the predicate frame, because every derivation of any other than the basic bioyaa involves automatically adding the features of number, gender, person and tense which, when the verb is in another gender, number, person or tense than 3rd. p. m. past, must consequently be deleted first. Such a procedure is methodologically undesirable in Functional Grammar. These were several theoretical reasons for not choosing actual words as lexical entries. The other alternative is to represent the roots separately. There is also empirical support for seeing the roots as separate entities: from child language and from mistakes made during language acquisition (of adults). This evidence, based on studies of Waiden (1982), Berman (1980, 1983), Berman and Sagi (1981), and Schwarzwald (1981) was presented in chapter 1. The next topic following from the previous discussion is the representation of the roots. The roots in the Semitic languages form a very interesting but complex topic in themselves, described among others in Bolozky (1971) and Greenberg (1950). In this study the discussion is limited to their representation in an FG lexicon either as independent entities or in other ways. The roots could be included in the lexicon as separate entities, but this would leave the problem of representing the derivational processes of the concrete verbs unsolved. FG permits us to list the roots in the predicate-frames as if they were a predicate. For example: (7)a b c

k.t.b. V-B1 (xl)Ag (x2)Go h.l.x. V-B1 (xl)Ag y.c.v. V-B3 (xl)Ag (x2)Go

'write' 'go, walk' 'stabilize'

48

Here the roots are seen as verbal predicates, and listed in the lexicon in their basic binyan. (The notion 'basic binyan' will be explained later.) This representation avoids the previous problem whereby number and gender were automatically assigned to the verb. However, it still faces the problem of category assignment: are the roots verbal, nominal or adjectival? It would appear in fact that roots have no category at all. If their representation is based on the notion of 'basic predicate' and predicate-frames, the isolated root does not have to be assigned a category at all. The details of how this should be done will be presented in sections 3.4 and 3.5 . 3.4 An F6 lexicon for Modern Hebrew The rest of this chapter will be on representation of the MH verbal system concerns the various components of the organization. I wish to emphasize that dictionary but a lexicon which is part

how FG envisages the in the lexicon. Section 3.4 lexicon, and section 3.5 its what is proposed here is not a of a linguistic model.

3.4.1 The root Of the four approaches to the lexicon surveyed in Schwarzwald (1975) this study adopts approach (ii), whereby all words should be represented together with their roots whenever the root is traceable. Many words in Modern Hebrew clearly exist as independent words, without being derived from a root. These are 1. those that entered MH as words, whether of biblical origin, or Aramaic or Greek origin. For example Sulban 'table', halon 'window', prozdor 'corridor' (from Aramaic) etc. 2. those that entered MH as loan words from which no root has been extracted such as koaliciya 'coalition', inflaciya 'inflation'. These may get Hebrew suffixes, for example in order 1 to form adjectives koaliciyoni 'coalitional' or inflaciyoni 'inflational . But they do not serve to form new roots, like telefoo —> t.l.p.n. (extracted root) —> tilpen 'phoned'. These words occur as basic predicates, usually nominal, and will be listed in the lexicon as predicate-frames similar to the basic predicate frames in say English or Dutch. All other Hebrew words consist of roots + patterns, which can be represented by the following table. Although it simplifies matters to some extent, 0 it is reasonably accurate: (8)

morphological component semantic value syntactic value morphological value

root

miSqal

bioyan

+

(+)

+ +

+

+

+

The morphological importance of the roots is best represented, in a system suggested by McCarthy (1981), where a root is seen as one noncontinuous morpheme. The binyanim system can be represented as follows:

49

(9)

root semantic value

+ morphological value

biayaa syntactic & semantic value

morphological value

The lexicon should clearly indicate that the actual verbal forms consist of these two separate components. Therefore, in the lexical entry for Hebrew there are no basic predicates of the type found in English or Dutch, since the roots have to be listed not only in a word but also independently. When discussing the representation of the roots, it should also be noted that in Modern Hebrew there are two types of roots: (i) primary; (ii) extracted, or secondary. The primary roots are simply all those which originally existed in the language. In addition, there is a comparatively small group of roots which are extracted from words. This process of 'root extraction' is described among others by Werner (1982). For example from the loan word telefoa 'telephone', the root t.l.p.n. is extracted from which in turn the verb tilpen 'phoned* Op.s.m. PIEL) is derived. Such a process can occur not only with loan words but also with original Hebrew words. For example: (10)

root 1 > N (original) h.1.1. s.f.r.

tehila 'beginning' mi spar 'number'

> root 2 (extracted) t.h.l. m.s.p.r.

>

V/N hitbil (B5) 'began' misper (B3) 'numbered*

This process of extracting a new root is fairly productive in Modern Hebrew, though it also occurred in Mishnaic and medieval Hebrew. Note that in most cases the extracted secondary root is quadri-literal. When the secondary root is not extracted from a loan word but from a Hebrew word which is itself derived from a root, it refers to a semantic field that is a hyponym of the semantic field referred to by the original primary root. Such processes of extracting a hyponymic secondary root occur mostly when there is a need to express a certain action, but the original root is 'used up', no longer available. In one of the examples above there are in fact three homophonous roots s.f.r. The one meaning 'count', occurs only in the PAAL; it is ( + t r ) , but means 'count', not 'number*. The FIEL and PUAL with s . f . r . is already occupied by siper 'told', and s . f . r . in PIEL and HITPAEL means 'cut someone's hair' and 'get one's hair cut 1 respectively. Therefore, the noun from s . f . r . , mi spar 'number', is taken as the basis of a new root, m.s.p.r., which then gets an active form in the PIEL·: misper 'numbered'. The semantic and morphological value and function of these secondary roots is the same as that of the primary roots. This, however, does not necessarily mean that they should get a representation in the lexicon similar to the primary roots. In fact, it seems more logical to represent them as derivatives. For example

50

(11)

s.f.r. binyania: safar (Bl) 'counted nispar (B2) 'was counted*

aiSqalia: mispar 'number' i mispur 'numbering' memuspar 'numbered*

m.s.p.r. i binyania misper (B3) 'numbered' muspar (B4) 'was numbered*

3.4.2 The patterns The morphological patterns, verbal and nominal, are also represented differently in the present study than in either the traditional or the other contemporary studies of Hebrew. The difference is that the binyania are conceived as having a typical argument structure and valency corresponding to the syntactic function of the biayan. Hence, the binyania resulting from valency reduction like the NIFAL, PUAL, HUFAL and HITPAEL have a typical predicate-frame with one argument only, which is Ag/Fo, Po or Proc. The binyania PIEL and HIFIL which are in most cases transitive or causative are typically associated with a predicate frame that has two or three arguments at least, with (xl)Ag/Fo and (x2)6o, and in the case of causative also an (xO)causer. 3.4.2.1 The aitqtlio The aiSqalia and the roots have only semantic and morphological value, whereas the binyania have syntactic as well as semantic and morphological value. The semantic value of the aiSqalia is in many cases not regular or consistent. Sometimes aiSqalia have a consistent meaning, but even that does not apply to all their occurrences. The maCCeCa pattern illustrated below, can have two meanings, which are consistent. But it is not possible to predict which meaning it will have with a certain root. (12)a maCCeCa = s.t.l. k.b.s. s.p.r. b maCCeCa = c.l.m. s.r.t. m.t.r.

'place of doing CCC' (CCC=root consonants) 'plant' —> aaStela 'greenhouse, nursery' 'wash clothes' —> aaxbesa 'laundry 1 'cut hair' —> aaspera 'hair dresser's' 'instrument for doing CCC' 'photograph* —> aacleaa 'camera 1 'film' —> aasreta 'film camera' 'rain' —> mamtera 'irrigation tap*

There are other aiSqalia which are no more than a morphological pattern consisting of affixes added to the root in order to form a word, but without adding any meaning of their own. For example (13)a b c d

CeCeC : even 'stone', elef 'thousand' CaCeC : haze 'bust', sade 'field 1 CaCiC : aviv 'spring', yahid Only, sole' CeCiCa: ooiya 'ship 1 , pniya 'turn*

Avineri (1976) lists about 120 aiSqalia, whereas according to Shalgi (1976) there are 243 aiSqalia. The consistency in the aiSqalim system is very low. Therefore, usually the roots + aiSqalia are listed as independent words. As mentioned already, the aiSqalia are much less

51

regular than the biayaaim. There are several groups where we get regularities in the miSqalim; the overall system, however, is open while that of the biayaaim is closed. λ schematic representation of the miSqalim and biayaaim in the lexicon would be: biayaaim

miSqalim

regular

80%

10-15%

irregular/ idiosyncratic

20%

85-90%

As we can see from this table, the biayaaim system is much more regular than the miSqalim system. Furthermore, the biayaaim system is closed in the sense that no new biayaaim enter the verbal system of MH. Rather, there is a tendency to simplify the verbal system by which only part of the biayaaim are productive (i.e. new verbs are formed with them). (See chapter 5, sections 5.3.1 and 5.6.) The miSqalim system on the other hand, is open; new miSqalim enter the system, (like maCCeCa for types of disease) or unfrequent ones become productive. Therefore, the miSqalim probably have to be simply listed individually. There are too many of them, and their occurrence6 and meaning is too unsystematic to allow any other representation. Therefore, I suggest listing the nominal forms simply following the root:

(14)

Jc.t.b. 'write iij.uyaLij.iu,

aij,3tfaj.j.ai.

predicate-schemes: katav 'journalist 1 CaCaC k.t.b. V-B1 mixtav 'letter' 'letter 11 miCCaC ktiv 'spelling CCiC (xl)Ag (x2)Go k.t.b. V-B2 maxteva 'desk' maCCeCa (xl)Proc ktovet 'address' CCoCet etc. katvanit 'typist'1 CaCaCit katvanut 'typing CaCaCnut haxtava 'dictation' haCCaCa ktuba 'marriage contract' CCuCa etc. (15)

s.l.h 'send' biayaaim: predicate-schemes: S.l.h. V-B1 (xl)Ag (x2Go etc.

misqalim: ndSloah 'cargo 1 miCCoaC miSlahat 'delegation' miCCaCat slihut 'mission 1 CCiCut

For a more detailed description of the miSqalim the interested reader should turn to Hir (1979) and Rosen (1977).

52

3.4.2.2 The binyaoia The bioyanim, on the other hand, are not only a morphological component of the concrete word, but also determine its syntactic features; in this sense they have a 'syntactic value'. This has been pointed out already in chapter 1. Often the binyanim are the only possible way of expressing certain syntactic and/or semantic features. This is always the case with NIFAL, PUAL and HUFAL as pseudo-passive, and often with HIFIL as causative, or HITPAEL as reflexive or reciprocal. Before proceeding to the actual representation of the binyanim in the lexicon, this is 1the appropriate place to expand on the notion 'predicate-scheme which was introduced in chapter 2 section 2.5 . A predicate-scheme is a general type of predicate-frame which holds for a certain binyan. This means that since each binyan has a prototypical valency, it can be associated with a 'prototypical predicate-frame 1 which would reflect this valency. The 'generalized predicate-scheme', introduced also in chapter 2 section 2.5 is a way of referring to a predicate-scheme without specifying the root. It is thus a somewhat more abstract notion. For example, the HIFIL- causative has the following generalized predicate-scheme: (16)

C.C.C. V B6 (xO)causer (xl)causee [(x2)Go]

a specific predicate-scheme would be (17)

n.p.l. V B6 (xO)causer (xl)causee

'made fall, throw over'

or the reflexive/reciprocal HITPAEL has the following generalized predicate-scheme: (18)

C.C.C. V B7 (xl)Ag/Pos

a specific instance of which would be the predicate-scheme (19)

r.b.c. V B7 (xl)Ag/Pos

'wash oneself

Thus the generalized predicate-scheme is an easy way of talking in an abstracted way about the properties of the binyanim only, whereas the predicate-scheme itself puts equal emphasis on the root too. For a concrete verb, the root contributes the meaning, and the binyan the category and the valency. The roots in the lexicon are represented in predicate-schemes. A predicate-scheme includes the following information: a. the root b. the category of the scheme (explained later) c. if verbal, the basic binyan d. argument slots: number, semantic function of the arguments and selection restrictions on the terms. For example: (20)

s.l.h. V-B1 (xl)Ag (x2)Go '.t.S. V-B7 (xl)Pos δ.p.c. V-B3 (xl)Ag (x2)Go

'send' 'sneeze' 'restore'

Of course, the roots themselves have ultimately no category. In this context too the category V is not assigned to the root, but to the predicate-scheme as a whole. In this the predicate-scheme differs from a

53

predicate-frame: in a predicate-frame the category refers to the predicate, whereas in the predicate-scheme it refers to the scheme. For the present study the verbal part of the grammar is chosen as the point of departure for organising the whole lexicon. Before going into the representation of the binyanim, it is necessary to clarify and justify the distinction between 'basic* and 'derived' binyanim, on which this representation is based. The distinction between 'basic 1 vs. 'derived' binyanim is made for the following reasons: a. methodological: since FG is the theoretical framework of this research, its descriptive mechanism is adopted, at least as a working hypothesis. This means that unless proven unsuitable, the FG descriptive mechanism incorporating the distinction 'basic' vs. 'derived', is used. b. Certain syntactic relations between binyanim can be explained more easily when one pattern is assumed to be basic. For example the HIFIL causative is derived from the PAAL (the latter being basic). The choice of the basic pattern will be explained and justified when discussing each concrete case. λ 'basic binyan' is defined as follows: The basic binyan for a certain root is that binyan from which all the other verbal forms of that root are derived. This is a morpho-syntactic operational definition. The criteria for establishing the basic binyan will be presented in chapter 4 section 4.2 and chapter 5 section 5.5 . Here they are introduced only briefly: a. semantically basic: which is the pattern that is the semantic basis for all the other patterns; b. morphological simplicity: the morphologically simplest pattern, i.e. the closest to the root; c. consistency with the rest of the system: if in the majority of cases a certain derivational line has been chosen on the basis of certain criteria, and in a particular case no other criteria can help to establish the basic pattern, the pattern which has been chosen as basic in the other cases will be retained. (This is in fact 'analogy'.) d. markedness: the PAAL (Bl) pattern is chosen as the unmarked pattern, because it can have both values: transitive and intransitive; when two patterns are otherwise equally basic, the PAAL is basic due to this criterion; The generalized predicate-schemes of all the binyanim are as follows: (21)

PAAL (+tr) (-tr) PIEL (c/+tr) HIFIL (+tr) NIFAL (-tr) (PP) HITPAEL (-tr) PUAL, HUFAL (pp)

c.c .c. c.c .c. c.c .c. c.c .c. c.c .c. c.c .c. c.c .c. c.c .c.

V-B1 {xl)Ag/Fo (x2)Go V-B1 (xl)Proc V-B3 (xl)Ag/Fo (x2)Go V-B5 (xl)Ag/Fo (x2)Go V-B2 (xl)Pos V-B2 (x2)Proc V-B7 (xl)Ag/Pos V-B4/B7 (x2)Proc

This type of generalized predicate-scheme results from the syntactic relation expressed by the binyan, regardless of the root. The root is a factor only in determining the occurrence or non-occurrence of a certain pattern with that root. Note that a root in the NIFAL, for example, will never have the same predicate-frame as in the PIEL or in the HIFIL·, and a verb in the HIFIL· will never have the same predicate-frame as in the HITPAEL.

54

3.5 The organization of the lexicon

In organizing the lexicon and representing the binyanim the following three options are possible: I. binyanim as predicate-schemes (22)

C.C.C. binyanim: C.C.C. Bl ( x l ) . - . ( x n ) [B2...B7]

A concrete entry would be as follows: (23)

k.t.b. binyanim k.t.b. V-B1 (xl)Ag (x2)Go

miSqalim: katav 'correspondent' katava 'newspaper article' mixtav 'letter' ktovet 'address' etc.

Here only the root in the basic binyan is represented as a predicateframe (it needn't be the PAAL· as in the example), followed by a list of all the other patterns in which that root occurs. This representation is similar in principle to Schwarzwald's and Berman's, except that it concerns all the roots and that the basic binyan does not have to be the PAAL. The main drawback of this organization is that the syntactic dependencies and the morphological derivations are not represented in any way; as they are not productive, they are not represented by predicate formation rules. Still, they too should be represented. Furthermore, this is an unorganized lexicon, in the sense that it is simply a list of all the roots existing in NH. II. binyanim in Configurational Sub-Systems (CSS): (24)

C.C.C. miSqalim:

binyanim:

Bl (xl)...(xn)

•L · · · ·

2. ... etc. CSS

Note that (24) includes all the binyanim in their prototypical functions and prototypical relations. A concrete example of a somewhat simpler entry would be as follows:

55

(25)a

k.t.b. 'write 1 aiSqalint: katav

'reporter'

mixatv 'letter*

ktiv 'spelling' etc.

binyanim (in a CSS): Bl (+tr) -v B2 (-tr)

The concrete predicate-frames associated with the binyaniot are as follows: (25)b Bl: katav V-B1 (xl)Ag (x2)Go B2: nixatv V-B2 (xl)Proc B5: hixtiv V-B5 (xO)Causer (xl)Causee (x2)Go B6: huxtav V-B6 (xl)Proc B7: hitkatev V-B7 (xl)Po According to this proposal the binyanim section will consist of the basic binyan of the root in the form of its predicate-frame. The other binyanim in which the root occurs are related to it via the CSS scheme applicable to the root. Compared to the previous proposal this alternative has the advantage of representing the syntactic and derivational relations between the binyanim. However, as a complete lexicon it is also an unorganized list of all the roots and the 'head' of an entry is a root. Note that both in I and II the representation of the binyanim and of the miSqalim is symmetrical, in that they are both derived directly from the root, and each lexical entry has two sub entries as it were, verbal and nominal, both on an equal level.

56

III. CSS as entry

(26)a

Bl (+tr)

-v B4(pp)

-v B6(pp)

The generalized predicate-schemes associated with this diagram are: (26)b

Bl: B2: B3: B4: B5: B6: B7:

C.C.C. C.C.C. C.C.C. C.C.C. C.C.C. C.C.C. C.C.C.

Roots:

V-B1 V-B2 V-B3 V-B4 V-B5 V-B6 V-B7

aiSqalia: 1

(xl)lg (x2)Go (xl)Proc (xl)Ag (x2)Go (xl)Proc (xO)Causer (xl)Causee t(x2)Go] (xl)Proc (xl)Po

([] = optional)

etc.

According to this suggestion a lexical entry is not a root but a Configurational Sub-System (CSS), organised as in (26). The concrete roots are featured not in the CSS schemes, but are listed afterwards. The CSS and the relations in it apply to all the roots individually. (Concrete examples of such entries are (28)- (3D). 'Configuration' refers to all the binyanim in which a certain root occurs; 'sub-system* to the group of roots sharing the same configuration. In other words, not all the roots are seen as one whole system, but each root or group of roots which occur in the same biayanim (i.e. share the same configuration) as members of one sub-system. (On the configurational sub-systems see also chapter 5 section 5.3.) The lexicon suggested here is based on distribution, which is a new approach to Hebrew. Its entries consist of the basic binyan and its predicate-scheme (and not root + bioyan) and of all the binyanim derived from the root (and their semantic and syntactic function), as well as of the valency operations for deriving them. This is followed by a list of roots which all share this CSS, and every root is followed by a list of the aiSqalim derived from it. The disadvantage of this approach is that possible regularities in the ntiSqalint system are not represented. This is, however, also the case in the other two alternatives, and furthermore, to the best of my knowledge such regularities are uncommon. Note that the representation of the nominal and the verbal patterns is not symmetrical, but this is not considered a problem. This representation has a very considerable advantage over the others in that it organizes the lexicon. It does not simply give a long list of all the roots, but organizes them in groups, according to a certain principle. In addition, this organizing principle which uses CSS's to express derivational relations between the binyanim as changes of valency, is also used in connection with the productive predicate formation rules. Thus, the whole description of the verbal system acquires a certain unity. Next to the regular cases there should be a list of the idiosyncracies, i.e. the cases when a certain verbal form does not have the meaning expected on the basis of the root + binyao. These may be complete idiosyncracies, or result from processes which were regular in

57

the past but which have undergone a change of meaning. Such are for example the so called 'lexicalized causatives', i.e. verbs in the HIFIL, but which in MH no longer express direct causativity. For example (27)

biskir 'to rent 1 (s.x.r. + HIFIL·) and not 'cause to hire' beSia 'accuse' (a.s.m. + HIFIL) and not 'cause to be guilty'

Consider the following CSS's as illustrations of this proposal. (Note that (28)-(31) is not an exhaustive list of the CSS's but merely some illustrations. The complete list is included in chapter 5.)

(28)a -v(2nd/irg.)

B2 (pp) The generalized predicate-schemes associated with these bioyanim are: (28)b B3: C.C.C. V-B3 (xl) Ag/Causee (x2)Go B2: C.C.C. V-B2 (xl)Proc B5: C.C.C. V-B5 (xO)Causer (xl)Causee (x2)Go B7: C.C.C. V-B7 (xl)Po

B4: C.C.C. V-B4 (xl)Proc Roots: 1. p.a.r. 'glory 1 , glorify' a. idiosyncracies: hitpaer V-B7 'boast' (lexicalized reflexive ?)

b. peer M (xl)0 'glory', (CeCC) tiferet N (xl)O'glory' , (ti+CCeCet) mefoar A (xl)0 'glorious* (meCCoaC) y.t.r. 'leave over' a. idiosyncracies: — b. niSqalim: yitra N (xl)0 'left over' (CCiCa) yitaron N (xl)0 'advantage' (CiCaCon) b.b.a. 'hide 1 a. idiosyncracies: — b. miSqalim: mahbo N (xl)0 'hiding place'

(maCCC )

y.c.v. 'put, stabilize' a. idiosyncracies: — b. miSqalim: 1 yecivut N (xl)0 'stability 1 yaciv A (xl)0 'stable S.r.S. 'root' a. idiosyncracies: —

(CeCiCut) (CaCiC)

58

miSgalim: SoreSV (xl)0 'roof (CoCeC) SorSi A (xl)0 'primeval' (CoCCi) muSraS A (xl)0 'rooted' (muCCaC) SorSiyut A (xl)0 'rootedness' (CoCCi+ut) (29) a Bl (-tr) B2 (i)

B3 (+tr) (iter)

The generalized predicate-schemes associated with these biayanint are: (29)b Bl: C.C.C. V-B1 (xl)Po B2: C.C.C. V-B2 (xl)Po B3: C.C.C. V-B3 (xl)Ag (x2)Go B5: C.C.C. V-B5 (xO)Causer (xl)Causee [{x2)Go] B7: C.C.C. V-B7: (xl)Po Roots: 1. g.a.h. 'pride' a. idiosyncracies: — b. miaqalia: staara N (xl)0 'pride' (CaCaCa) gee A (xl)0 'proud' (CeCeC) yaavtao A (xl)0 'proud* (CaCaCan) 2. c.v.h. 'scream' a. idiosyncracies: hictaveab (B7) 'scream + sudden' b. aiaqalia: cvaba N (xl)0 'a scream'

(CCaCa)

(30)a

B7 (i)

The generalized predicate-schemes associated with these binyanim are: (30)b Bl: C.C.C. V-B1 (xl)Po B2: C.C.C. V-B2 (xl)Po B5: C.C.C. V-B5 (xl)Causer (x2)Causee B7: C.C.C. V-B7 (xl)Po Roots: 1. p.g.m. 'harm* a. idiosyncracies: — b. miSqalim: pagum A (xl)0 'harmed' (CaCuC) pgatt Η (xl)0 'fault' (CCaC)

59

2. c.r.d. 'hoarse' a. idiosyncracies: — b. aiSqalia: carud λ (xl)0 'hoarse' (CaCuC) ceridut N (xl)0 'hoarseness* (CeCiCut)

(31)a

Bl (-tr) ^^ B7(-tr)

The generalized predicate-schemes associated with these bioyaoia are: (31)b Bl: C.C.C. V-B1 (xl)Po B3: C.C.C. V-B3 (xl)Ag (x2)Go

B5: C.C.C. V-B5 (xl) Causer (x2)Causee B6: C.C.C. V-B6 (xl)Proc B7: C.C.C. V-B7 (xl)Po

Roots :

1. g.n.n. 'protect* a. idiosyncracies: hitgonen (B7) 'defend oneself* (ref) b. miSqalia: magea Ν (xl)0 'shield* (maCCeC) auyan λ (xl)0 'protected* (muCCaC) nageo Ν (xl)0 'protector* (maCCeC) 2. ' . u . f . ' f l y * a. idiosyncracies: — b. aiSqalia: aa'of Ν (xl)0 'flying' 'afifon Η (xl)0 'kyte'

(maCCoC) (CaCiCon)

3. r.u.c. 'run* a. idiosyncracies: — b. miSqalim·,'1 rac Ν (xl)0 'runner' (CaC) meroc Ν (xl)0 'competition 1 (meCoC) rica N (xl)0 'running' (CiCa) 4. n.h.r. 'flow 1 a. idiosyncracies: —

b. nabar N (xl)0 'river' (CaCaC) nebira N (xl)0 'flowing' (CeCiCa)

60

5. e.h. k. 'laugh' a. idiosyncracies: — b. niSqalim: cbokV (xl)0 'laughter 1 (CCoC) aacbik A (xl)0 'funny' (maCCiC) cibkuk H {xl)0 'gigle' (CiCuCuC)

6. n. g. S. 'go to' a. idiosyncracies: — b. miSqalimi nayaS N (xl)0 'tray 1

(maCCaC)

7. n. s. n. 'breath' a. idiosyncracies: — b. ßiiqalia: aeSiaa N (xl)0 'breath' (neCCiCa) aiSum H (xl}0 'breathing' (niCCuC)

8. n.u.d. 'move* a. idiosyncracies: — b. miSqalim noded N (xl)0 'nomad' (CoCeC) aedida N (xl)0 'migration' (CeCiCa) 9. s.y.t. 'sail* a. idiosyncracies: —

b. Sayat N (xl)0 'sailor' (CaCaC) Sayatut N (xl)0 'sailing' (CaCaC+ut) Sayit N (xl)0 'sailing1 (CaCiC) aaSot N (xl)0 Oar' (maCoCC) As mentioned, this is not an exhaustive list, but merely some illustrations of the more complex CSS 's. If adopted, this proposal would result in the following general lexicon: (32)

l e x i c o n predicates:

terms:

B1 CSS1//-B2

basic:

roots: Rl, R2, etc

pronouns

migqalia: . . . B2 /U 5 "x .

derived: nouns and adjectives

B6

etc roots: Rl, R2, etc miSqalia; . . .

This suggestion deviates from the lexicon proposed in Dik (1981, 1983) and other studies in FG in that the lexicon includes derived predicate

61

frames besides basic ones. This seems, however, to be the only logical place for these derived patterns which do not result from productive processes. Another point in which it deviates from the lexicon proposed in Dik (1981, 1983) is the inclusion of elements which are not stems: the roots and the binyania. Dik (1979) argues against abstract meta-linguistic elements in the lexicon. The roots, however, are not abstract in this sense, because they do have morphological, semantic (and psychological) reality, and they do occur concretely in the words derived from them. 3.6 Summary In conclusion, the F6 lexicon presented in this chapter for a Semitic language differs in some respects from the lexicon for an Indo-European language. The differences affecting FG concern its components and its organization. But its basic principles remain unchanged. A lexicon for a Semitic language, here illustrated by Modern Hebrew, will consist of roots in basic predicate-like representations, i.e. in predicate-schemes. The predicate-schemes, as mentioned, represent valency values typical of binyania when expressing a syntactic-semantic value (+/- transitive, causative, reflexive, etc). Thus there are prototypical predicate-schemes for the various biDyaoia. The lexical entries are not basic predicates, as in say English or Dutch, but CSS's, consisting of predicate-schemes, the roots that belong to it and the aiSqalia derived from each root. A basic predicate is also a schematic representation of the syntactic and semantic features of a content word. Only, the actual word is not filled in concretely, but is represented by a root. In addition to the 'basic predicate-frames', here the equivalent of the 'basic biayaa', the lexicon for HH also features those other bioyanim in which the root occurs and which are derived from the basic binyaa by non-productive processes. Consequently, such a lexical entry contains much more information than a conventional lexical entry for Indo-European languages. Overall, the lexicon presented here deviates considerably from those suggested so far in studies based on F6: it is not simply a list of all the content words of Modern Hebrew. For its entries are based on a syntactic, semantic and morphological principle (the CSS's) and each entry may include more than one root. The result is a more economical lexicon organized along lines similar to the predicate formation mechanism, namely processes of valency change. Thus a certain unity between the lexicon and the predicate formation mechanism in the biayaoia system is obtained. At the same time, the organization of the lexicon in configurational sub-systems, i.e. presenting a lexicon based on distribution, differs from other approaches to the lexicon of MB.

Chapter 4

Predicate formation rules: the productive processes in the binyanim system 4.0 Introduction1 The use of the notion 'valency' in connection with verbs is not new, but has been increasing lately. Allerton (1982), for example, uses it to describe the English verb. Dik (1985) discusses the predicate-frame and some of the relations between basic and derived predicate-frames in terms of valency. In FG the constituents which can be regarded in terms of valency are the predicates (verbal or nominal). The valency of a predicate is indicated by the argument slots in its frame. Dik (ibid, p.85) defines the valency of a predicate as the structure which is reflected in its frame. A predicate has a. quantitative valency: the number of arguments it takes; b. qualitative valency: the semantic features of the arguments (reflected in the selection restrictions) and the semantic functions of the arguments. Consider the following example (from Dik, 1985:97): (Da b

eat V (xl: (xl))Ag (x2:(x2))Go drink V (xl:(xl))Ag (x2:(x2))Go

These two predicates have the same quantitative valency (two) but their qualitative valencies differ as the Go is subject to different selection restrictions. Predicate formation rules can be seen as changing the valency of a predicate. Valency change can consist of increase or decrease of valency (quantitative changes); or of changes in the arguments concerning their semantic function or the selection restrictions (qualitative changes). The following types of changes in quantitative valency can occur: a. increase of valency: transitivization, causative formation b. decrease/reduction of valency: (i) first argument reduction- pseudo-passive formation (ii) second argument reduction- detransitivization, formation of verbs in de-actualized or generic constructions, and in reflexive and reciprocal constructions. With causative predicates the number of arguments is increased, due to the addition of an (xO)causer. In the case of valency reduction the reduced argument can be either the (xl) or the (x2). The former results in a pseudo-passive construction; the latter in a generic, reflexive or reciprocal construction. Both types of valency change can be found in the HH binyaoim system; I will argue that all syntactic relations as well as morphological and derivational relations between the binyanim in the verbal system can be described in terms of valency change (quantitative or qualitative). In fact a similar idea regarding the verbal systems of Arabic and Hebrew is proposed in Saad and Bolozky (1984) : ...basically all morphological verb processes in the two languages [Arabic and Hebrew] may be characterized

63

as transitivization or detransitivization phenomena, manifest as either a change in the number of arguments involved (increase or decrease) or change in the degree of semantic transitivity (strengthening or weakening) or in both. (p. 31) Before going on to specific manifestations of valency changes in the MB verbal system, the reader is reminded that the basic predicate-frame in the MB lexicon differs somewhat from the one in an English or Dutch lexicon (as suggested in Dik, 1983, 1985). The lexical entries are configurational sub-systems (CSS's), consisting of roots with the basic pattern and the derived ones. A root +binyan form an actual word, which can have a predicate-frame, as (2)a

Salah V-B1 (xl:(xl))Ag (x2)Go

A predicate-scheme on the other hand, reflects the syntactic-semantic environment typically expressed by a certain root + binyan and not by a concrete verbal form derived from the root. For (2a) it is (2)b

S.I.h -Bl (xl)Ag (x2)Go

A generalized predicate-scheme is a yet more abstract way of refering only to the valency structure of a binyan. For example {C.C.C. = root consonants): (3)a b c d

C.C.C. C.C.C. C.C.C. C.C.C.

V-B1 (xl)Ag (x2)Go ; C.C.C. V-B1 (xl)Ag V-B2 (xl)Ag V-B3 (xl)Ag (x2)Go V-B5 (xO)Causer (xl)Causee [(x2)Go]

(3a) shows transitive and intransitive generalized predicate-schemes typical of the PAAL; (3b) the intransitive generalized predicate-scheme typical of the NIFAL; (3c) and (3d) the transitive and causative generalized predicate-schemes typical of the PIEL and BIFIL respectively. The FG lexicon of Modern Hebrew consists of predicate-schemes as (2b) and not of predicate-frames as (2a), as is the case in lexicons for Indo-European languages (See also chapter 3). They remain unchanged regardless of the concrete root inserted. They have the following structure: (4)

pred V = root + binyan

The bioyania have a semantic, syntactic and morphological value of their own, independent of the semantic and morphological value of the root. The two together form the final syntactic, semantic and morphological features of the verbal form. The notion of valency has been used in various studies in FG (Dik 1985, Mackenzie 1983, Van Schaaik 1985) but so far it has been applied only to concrete predicates, either verbal or nominal. In this study the idea of valency change is connected to the biayanim and not to the verbal predicates, and is applied not to predicate-frames, but to predicate-schemes. This is new to FG. This chapter seeks to demonstrate how the syntactic and derivational relationships between the various biayanim can be described in terms of quantitative valency changes. Of course, all the concrete predicates which occur in the described binyanim should and do fit the same description. The major interest here, however, is in the syntactic and semantic aspects of the biayanim.

64

This chapter will deal with cases that are considered productive (as defined in chapter 2 section 2.4). They involve the major functions of the binyaaim, as listed below: (5)

PAAL (Bl)

+tr: C.C.C. V-Bl(xl)Ag (x2)Go; -tr: C.C.C. V-B1 (xl)Ag NIFAL (B2) -tr: C.C.C. V-B2 (Xl)Ag pseudo-passive : C.C.C. V-B2 (x2)Proc FIEL (B3) +tr: C.C.C. V-B3 (xl)Ag (x2)Go PUAL (B4) pseudo-passive: C.C.C. V-B4 (x2)Proc HIFIL (B5) causative: C.C.C. V-B5 (xO)causer (xl)causee (x2)6o HUFAL (B6) pseudo-passive: C.C.C. V-B6 (x2)Proc HITPAEL (B7): -tr: C.C.C. V-B7 {xl)Ag/Pos

As mentioned in chapter 1, the binyania can have other functions too, which would be expressed in other predicate-schemes. The central hypothesis of this chapter and the next is that the syntactic and derivational relationships between the bioyanim can be described as changes of valency. In chapter 2 section 2.4 the derivation within the verbal system of KB have been derived into two types: productive predicate formation rules and non-productive regularities. At this point it is possible to raise the following objection against this division into two groups: as all the roots are included in CSS's in effect all the processes involving the roots apply to the CSS's. Hence, all the processes termed productive here also take place in and between CSS's, and should be formulated as such. This is, of course, true. All processes involving roots involve the CSS's too. Nevertheless, the distinction made here between the productive rules and the non-productive regularities is justified. Methodologically, because it is part of the FG model. Empirically, because several studies (Bolozky 1983, Berman 1983, Berman & Sagi 1982) show that the binyania with the functions listed in (5) are productive among speakers of Hebrew, both children and adults. Therefore, the processes by which these binyaaim are derived are productive too, lending them a general validity. Furthermore, the CSS's represent a distributional level. They are introduced in order to capture whatever regularities may exist in addition to the general processes. If all the rules were to be stated as rules operating in and between CSS's instead of as predicate formation rules and regularities, the CSS's would replace the binyania as the central organizational principle of the MB verbal system. The aim of the present study, however, is to describe the syntactic-semantic functions and processes of the binyania. Their distribution is only a secondary aid to help find and formulate these functions and processes. 4.1 Transitivity in Hebrew 4.1.1. Transitivity or valency Since the notions valency and valency change are closely related to transitivity (and transitivization /detransitivization), these terms and their use here are clarified first. An additional reason for doing this is that the term 'transitive verb 1 is not used consistently in the various studies of MH.

65

Saad and Bolozky (1980, 1984) look at the subject of transitivity (in Hebrew and Arabic). Their definition of transitivity is purely semantic, and it is not polar but scalar (following Hopper & Thompson, 1980). [transitivity] we believe should be restricted to two features only, one syntactic the other semantic: the number of participants involved and the degree of affectedness of the object (or degree of 'patienthood'). Thus a highly transitive verb involves at least two participants and the patient is highly affected. Saad & Bolozky (1984) illustrate the ascending degree of transitivity of the same verb in different contexts by the following sentences: (6)a b c d

John John John John

ate ate ate ate

his food, lots of food, spaghetti, poisoned food.

Although the verb is the same, due to the situation described 'John' in (6d) is, according to Saad & Bolozky, more affected than in (6a). The reason is that eating poisoned food affects one more than eating normal food. This reason, however, is not linguistic, but due to the situation referred to. This is a very vague criterion, and it is rather difficult to use it consistently. Furthermore, in (6a) to (6d) two criteria are mixed through each other. In (6c) to (6c) the object of 'eat* is increasingly more specified, but in (6d) there is suddenly another criterion: the Agent 'John' is more affected than in (6a) - (6c), but the Object of 'eat' is in fact less specified than in (6c). Hence, although 'transitivity* is certainly a semantic notion too, this purely semantic definition used by Saad & Bolozky remains too vague when it comes to concrete cases. Another definition of transitivity, based entirely on syntactic and formal features of the verbs is presented in Stern (1979). Stern limits 'real transitives' to those verbs that meet both criteria listed below: a. have an obligatory (x2) which is preceded by et when it is [•»•definite]; b. can undergo 'passive transformation', i.e. can occur in passive constructions. Criterion (a) is clear and straightforward. In criterion (b), however, the use of the term 'passive' needs clarification. Stern calls a construction passive if it has a verb in the NIFAL, PUAL or HUFAL pattern (even without Ag), as long as it is not statal. In the examples below, Stern considers only (b) as passive (he calls it 'Vorgangspassiv') and (c) as 'statal passive' (Zustandspassiv). (7)a b c

(8)a

hamore patah (Bl) et hadelet. the-teacher opened 3p.s.m. GN the-door 'The teacher opened the door', hadelet nifteha(B2). the-door was-opened 3p.s.f. 1 The door opened'. hadelet ptuha (PAUL·) 2 the-door open-fern. the-door is-open 'The door is open' hem maxru (Bl) et habayit Selahem mizman. they sold 3p.pl.m. CM the-house the-theirs long-ago 'They sold their house long ago'.

66

b c

habayit haze nimkar (B2) mizman. the-house the-this was-sold long ago 'This house was sold long ago*, habayit nimkar (beaoni). the-house is-sold 'The house is sold'.

In other words, Stern considers the benoai of the NIFAL, PUAL and HUFAL patterns and the benoni of PAUL to be statal passives. The 'real 1 passives are expressed by the same patterns, but in the past or future tenses. In other words, benooi constructions are seen as 'statal' and therefore not 'real passives'. Another view, which is also followed here, connects transitivity to valency, i.e. to the number of arguments. Before going on to apply valency to the binyaaim, it is necessary to state clearly how the notion of transitivity is used here. When looking at transitivity in the verbal system of MH, there are some clear cases, but also many less clear ones. On the whole MH verbs can be divided into the following groups: et + (x2)Go a. obligatory (x2) oblig. marker* (x2)Compl (i)

valency 23

\ b. optional/ reducible (x2) (ii)

valency 1: intransitive verbs (these are irrelevant to this discussion) (iii) valency 3: a. basic b. derived: HIFIL causative (These verbs will be treated in section 4.5}

ad (i) The clear cases are those where the verb has two arguments, and (x2) is preceded by the particle et. (This particle occurs only with a +definite nominal.) For example (9)

yosi kvar kara (Bl) et hasefer haze. Yosi already read-3p.s.m. GM the-book the-this 'Yosi already read this book'.

The particle et is usually translated as 'accusative marker'. However, since it occurs only before an (x2) which is Goal, it can be seen as 4 'Goal Marker* (henceforth GM). (An indefinite Go has no explicit mark.) Somewhat more problematic are those verbs which are definitely 2-place verbs, i.e. have an obligatory (x2), but take another particle, never et. (As listed for example in Berman (1978) and Rosen (1977)). These are illustrated by the following examples, where the verb is always transitive and the particle is obligatory: (10)a

david 'azar (Bl) lerina. David helped -3p.s.m. to-Rina 'David helped Rina'.

67

b c (ll)a b c

*david 'azar (Bl) et rina. David helped-3p.s.m. GM Rina ?david 'azar (Bl) (unclear without a context) David helped david hiSpi'a (B5) 'al havero. David influenced- 3p.s.m on friend-bis 'David influenced his friend* «david hiSpi'a (B5) et bavero David influenced-3p.s.m. GM friend-bis ?david hiSpi'a (B5) (unclear without context) David influenced

Rosen (1977:108) further demonstrates the obligatory character of these clitics or particles by showing that in the case of these verbs omitting the clitic will change the meaning of the verb, as in (12)

la'aaod 'al 'to insist' lebacbi'a 'al 'to point out' lexabed be 'to offer 1

la'amod 'to stand' lebacbi'a 'to vote' lexabed 'to respect*

These verbs are similar to prepositional verbs in English, like 'call' vs. 'call forth*. The obligatory clitics are: be- 'in, with', me- 'from, le- 'to* and ke- 'as', and the particle 'al On'.

(ii) The second group consists of those verbs that take an optional et. According to Stern these too have valency 2; however, when these verbs occur with another particle they change meaning. Therefore, when they occur with one argument only, they have undegone valency reduction. Consider the following examples: ad

(13)a b

ha'as kirsem (B3) tbe-moth chewed- 3p.s.m. 'The moth chewed holes in ha'as kirsem (B3) the-moth chewed-3p.s.m. 'The moth chewed holes in

et hasveder GN the-jumper the jumper' basveder. in-the-jumper the jumper'

According to Stern (13 a,b) are equivalents, but in actual usage (13b) is less common and more literary than (13a). In addition, there is a difference in meaning: (13a) means that the moth chewed the whole jumper to pieces, whereas (13b) means that the moth chewed one or several holes in the jumper, i.e. completive vs. partitive. Instead of et + (x2) an infinitival construction too is possible, as illustrated below: (14)a b

rina Rina 'Rina rina Rina 'Rina

hifsika (Bl) et 'avodata lereg'a. stopped- Ip.s.f GH her-work for-a-moment stopped her work for a moment' hifsika la'avod (inf) lereg'a. stopped to-work for-a-moment stopped working for a moment'.

Here (14b) is much more frequent than (14a); the difference in meaning is more subtle than in the previous pair, and it is not directly obvious.0

68

Still, the choice of using et or an infinitival form is not arbitrary and the et is therefore not optional, and the verb in (14b) is in fact reduced, i.e. has valency 1. According to Stern there is a third type of verbs which also belong to this category, namely those with an optional (x2). When the (x2) occurs, it is never Go (because it is never preceded by et). For example; (15)a

david pahad (Bl) mehakelev David feared 3p.s.m. from-the-dog 'David was afraid of the dog* *david pahad (Bl) et hakelev david feared 3p.s.m. GM the-dog david pahad. (Bl) david feared 3p.s.m. 'David was afraid'.

b c

Here too, however, there is a difference between (15a) and (15c); (15c) is in fact a detransitivized predicate derived from (15a). (Or rather would be, if it were used). (15c) is more generic or habitual than (15a); (15a) specifies an instance of (15c), as illustrated below: (16)a

b

?ani betuha seim nelex (Bl) levaker et rina, I certain-f. that-if go-lp.pl.fut to-visit GM Rina david yefahed (Bl) david afraid- 3p.s.m.-fut Ί am certain that if we will go to visit Rina, David will be afraid 1 . ani betuha seim nelex (Bl) levaker et rina I certain-lp.s.f. that-if go-lp.pl.fut to-visit GM Rina david yefahed (Bl) mehakelev. David afraid 3p.s.m.-fut from-the-dog Ί am certain that if we will go to visit Rina, David will be afraid of the dog.'

In other words, when referring to a specific event, pabad 'was afraid' cannot be used without an (x2), because the context requires it. (16a) would make sense only if the previous context explained why David is afraid. Hence it can be concluded that: a. (x2) in (16b) is Reason, not Go; b. verbs of this type are not synonymous with valency 1 or valency 2; when used with one argument only they receive a more generic interpretation. This fits one of the categories of detransitivization, as we shall see in 4.6.2 . In the examples listed above it was demonstrated that group (Ib) differs from group (la). The question now is whether an (x2) which is preceded by an obligatory particle other than et, so that the particle belongs to the verb should be assigned the semantic function Go or not? (The representation of the particle itself is not a problem: it will appear in the lexicon, together with the verb in question.) 6 Berman (1978b), like Stern, treats all these verbs as +transitive and allows the Go to be preceded by another particle than et. She suggests listing these verbs in the lexicon accompanied by the particle they take. This seems to me an acceptable solution for the particle. But the important point here is the semantic function of this obligatory (x2). For the sake of clarity and consistency in this study the following definition is used:

69

when (x2) is preceded by et it is Go, and when it is preceded by any other clitic or particle, it is Compl(ement). This clitic or particle will be listed in the lexicon, together with the verb. Hence, both verbs with (x2)Go and (x2)Compl are considered two-place verbs whose predicate-frame is bivalent, λ two-place predicate-frame is therefore as follows: (17)

pred V (xl)Ag/Fo (x2)Go/Compl

An important implication of this definition is that the crucial feature for transitivity is the predicate-frame of the verb, and not the passivizability of the construction. This is preferable because in MB there are practically no passives in the proper FG sense of Subject assignment to Go. (See also section 4.6.1 .)In addition, this definition is not based so much on semantics, as in Saad & Bolozky (1984), but includes formal syntactic aspects of the verb. 7 It is, therefore, more clear cut and easier to use in concrete cases. Note that in this definition in fact the emphasis is put on the quantitative valency of the verbal predicate. This results in a criterion which is clear cut and easy to use and on which later derivational processes can be based. Transitivity, on the other hand, is more of a scalar notion, where semantic and textual factors also play a role. By using valency, it is possible to avoid the problems concerning the definition of transitivity for MB, encountered for example in Herman (1979) and Stern (1979). flere it is enough to state that the following predicate-frame is accessible to valency reduction (of the detransitivization type): pred V (xl)Ag/Fo (x2)Go/Compl. Bence, whenever a biayan has a predicate-scheme based on this type of predicate-frame, it can undergo valency reduction. The following points about (17) should be noted: 1. (x2) is not necessarily Go. It can be also Compl as in the verbs that have an obligatory (x2) but do not take er. 2. only the Go is sometimes accessible to Subj assigment; the Comp is never accessible to Subj assignemnt. 3. This definition is based on the predicate-frame of a verb and not on semantic features (like 'affectedness') or syntactic features (like passivizability and cliticization). Another reason to prefer valency to transitivity is that transitivity is a notion which, at least in Hebrew, is connected with specific verbs and specific constructions. Valency, on the other hand, is a more general feature which can be associated with a binyan without filling in a concrete root. Consider the following examples: (18)a b

bahodeS haaharon hamehirim hityacvu (B7). in-the-month the-last the-prices stabilized 'In the last month the prices got stable.' habayalim hityacvu (B7) bamisdar. the-soldiers came/ were present in-the-roll-call 'The soldiers came to the roll-call'

The verb in both (18a,b) is the root y.c.v. 'stable' in the BITPAEL pattern; yet in (18a) it has an inchoative meaning, and in (18b) an active-intransitive meaning. That is to say, the precise meaning (or function) can change, depending on the construction i.e. the arguments. The valency, however, is in both cases 1; nor did I encounter examples

70

where a root in a certain binyaa had in two different contexts a different valency. 4.1.2 Productivity The term 'productive* is used in this chapter according to the definition in chapter 2 section 2.4 . The processes described in this chapter as predicate formation rules are considered to be productive. Those processes which show a clear regularity but apply to a finite group of elements only, and thus form a list of cases (though the list might be large), are considered unproductive and will be treated in chapter 5. 4.2 Basic and derived binytnim This section will deal with the distinction between basic and derived patterns, the justification for this distinction and the criteria used in establishing it. There are two ways of looking at the verbs in MH: a. each one may be seen as derived directly from the root; b. one (or more) of such root + binyan combination(s) may be seen as basic, from which the rest are derived. In the former case there is a total symmetry between all the verbal forms, but not in the latter. In the latter case some binyanim are derived from another binyaa (predicate), and not from the root. In chapter 3 approach (b) has been adopted. A basic binyan was defined as follows: the basic biayao is that binyan from which all the other verbal forms of that root are derived. Note the following two points about this definition: 1. There is not one general basic pattern for the whole verbal system; it can vary depending on the root. 2. This definition applies primarily to morpho-syntactic derivational processes. The criteria for establishing which binyan is basic were presented in chapter 3 section 3.4.2.2 . They are repeated here only in short: a. semantic simplicity: which pattern forms the semantic basis for the others. Some verbal forms are semantically more complex than others. For example, a causative or a reflexive form is semantically more complex than an active transitive or intransitive. b. morphological simplicity: the closer a verb form is to the root, the simpler it is morphologically. Thus the PAAL, PIEL, and PUAL patterns consist of the root consonants only, with infixed vowels, whereas in the NIFAL, HIFIL, HUFAL and HITPAEL there are also prefixes: /ni/, /hi/, /hu/ and /hit/ respectively. According to this criterion the PAAL, PIEL· and PUAL· are more basic than the NIFAL·, HIFIL, HUFAL and HITPAEL·. c. markedness: the PAAL· (Bl) is chosen as the unmarked binyan because it can be both transitive and intransitive; when two candidates for the basic binyan are equal on all other criteria, and one of them is the PAAL·, the PAAL· is chosen as basic. d. consistency with the rest of the system: if in a specific case two alternatives are equally basic and one of them has been chosen in other cases as basic, then that pattern will be the basic one again. Apart from the fact that it fits in with the FG distinction between basic and derived predicates, this distinction between basic and derived binyanim can be justified also by the existence of distributional regularities in the occurrence and function of the binyanim; these

71

regularities can be described most efficiently as being derived from a basic binyan. For example, the occurrence of a root in the HIFIL pattern as causative depends on the occurrence of the same root in the PAAL pattern; in such a case it seems more efficient to derive the HIFIL from the PAAL, than to derive both from the root (which is the only other alternative). Another example is the pseudo-passive binyaaim PUAL and HUFAL, whose occurrence depends on the occurrence of the same root in the PIEL and HIFIL·. It is possible to raise against this that it is simply a question of co-occurrence without involving any further relationship between the bioyaaia and that derivation involves a stronger type of relationship. This further step from co-ocurrence to derivational relationship is made on the basis of the FG predicate formation mechanism. 4.3 The basic bioytaim KB has two groups of binyaniat which are basic: productive and unproductive. The productive group consists of patterns which still admit new verbs to the system. These are the PIEL·, HIFIL·, and to some extent HITPAEL. The NIFAL is unproductive as a basic pattern, i.e. no new verbs enter the system in this pattern. The formerly productive PAAL· is the basic pattern of a large group of roots. In contemporary Hebrew, however, new verbs entering the language mostly follow the PIEL· or HIFIL· pattern. A small group of intransitive verbs follows the HITPAEL·. Most verbs entering MH nowadays are denominal. According to Bolozky (1978) there are several factors determining which basic pattern these new verbs will adopt. The following factors play a role: - pronounceability: quadriliteral or longer roots will be in the PIEL because in the HIFIL· they form an unpronounceable word. - whether the biayaa in question is typically transitive or intransitive. - occupied slots. Sometimes a denominal verb cannot be realized in a certain biayaa because it is occupied already by a homophonous root. For example (20)a b

paSut 'simple' —> piSet (B3) 'simplified' -/-> bifSit (B6) 'undressed' (is not derived from paSut) matuo 'moderate' —> mi tea (B3) 'caused to be moderate' -/-> bimtia (B6) 'waited 1 (is not derived from «arun)

In addition to these main factors there are factors which play a more incidental role, such as: phonological restrictions, analogies in the formation of a verb, and the need to preserve the structure of the noun/adjective. Bolozky (ibid) even suggests a hierarchy for these factors. The interested reader is referred to Bolozky (1978). The other pattern which can be basic is the HITPAEL. As a basic pattern the HITPAEL is always intransitive and often also the only verbal pattern in which the root in question occurs. For example (21)a

b c d e

bit'ateS 'sneezed' biSta'el 'coughed' bitpalmes 'argued' histakel 'looked' bitrakeab 'argued'

(from (from (from (from (from

'.t.s.) s.'.l.) p.l.m.s.) s.k.l) v.x.h.= y.x.h)

72

According to Barkali (1980) there are 42 roots which occur exclusively in the HITPAEL. Only 11 of these are used in MH. Three of these also occur in other binyanim; in these cases the HITPAEL is not basic. Of the regaining 8 verbs tiro are inchoative: hitbala 'became discoloured' and biStahrer 'became dizzy*. This leaves only six verbs using the HITPAEL as their basic and only pattern. Barkali (1980) also states that there are 20 roots listed as occurring only in the NIFAL. However, when they were also checked in Even-Shoshan (1967), only the following two were found: (22)a b

nikmaz 'shrank 1 (from k.m.z.) (Talmudic) aisJcad 'was tied' (from s.k.d.) (Biblical)

Neither of these forms is used in HH. None of the other 18 roots are listed in Even-Shoshan's dictionary, nor are any of them known to me. There are also Configurational Sub-Systems (CSS's) featuring more than one pattern, where the NIFAL or the HITPAEL could be basic. These are: (23)a

b

NIFAL:HITPAEL 1 root: m.h.h. ninba 'got crumpled' (according to Even Shoshan no HITPAEL) NIFAL:PUAL:HITPAEL 1 root: x.l.h 'waste' (according to Even-Shoshan also in PAAL, and as far as I know occurring only as xala (Bl) 'got wasted') HITPAEL:HUFAL 1 root: g.n.s. '? ' (does not occur in Even-Shoshan and is not known to me)

Note that more roots have the PAAL, PIEL or HIFIL as their only binyan: PAAL: 117 FIEL: 204 HIFIL: 30 The basic binyao, whether productive or non-productive, will be listed in the lexicon for all the roots. As mentioned in chapter 3, the entry is a CSS whose starting point is the predicate-scheme of the basic biayan, followed by all the other binyanim in which a certain group of roots occur. 4.4 The derived binyanim There are two binyanim which are always derived: the PUAL and the HUFAL, derived from the PIEL· and the HIFIL· respectively. Barkali (1980) would seem to contradict this, mentioning 13 roots that occur only in the PUAL, and 1 root that occurs only in the HUFAL. These forms, however, are not listed in Even-Shoshan, and are not known to me either. I have therefore chosen to ignore them, concluding that these two binyanim are in fact derived. The rest of this section will be about the derivational processes involved in deriving the binyanim. By examining the various concrete uses of these binyanim, I shall try to determine to what extent these can be described in terms of operations of valency change. Here the results are given before the examples on which they are based, so as to make the presentation more coherent. The data on which this description is based will be presented in sections 4.5 and 4.6 .

73

4.4.1 Patterns that can be both basic and derived The following patterns can be both basic and derived (not, of course, with the same root): a. PIEL and HIFIL, which if derived result from valency increase, mostly of PAAL·; b. NIFAL and HITPAEL, which result from valency decrease, mostly of P AAL. The PIEL is fairly often the transitive pattern of a root which is intransitive in the PAAL. This is valency increase, or transitivization, of the PAAL·. The HIFIL is very often the causative of a root which is mostly intransitive (but sometimes transitive) in PAAL·. These too, are cases of valency increase of the PAAL·. (The HIFIL can also express inchoativeness, in which case it has valency 1 (i.e. intransitive); these cases are too rare to be called productive, and will be dealt with in chapter 5.) The NIFAL and HITPAEL always have valency 1. The NIFAL is often the pseudo-passive of a root which is bivalent (transitive) in the PAAL·. It is derived from the PAAL· by first argument reduction. The HITPAEL· is the reflexive or reciprocal of roots which are bivalent (transitive) in the PAAL· (or PIEL· or HIFIL·), and is derived by second argument reduction. Reflexive and reciprocal are seen in some approaches (eg. the T6 approach) as bivalent. The reason why in F6 they are treated as univalent is discussed in section 4.6.2 . The HITPAEL is also active-intransitive, being derived by second argument reduction.

4.5 Valency increase Valency increase can be of two types: a. transitivization b. causativization In both cases the input is the PAAL pattern. Transitivization applies to the derivation of a transitive PIEL from an intransitive PAAL·, as in (24)

Saxax 'be quiet'

(25)a b

-

Sixex 'make quiet' (+tr) (referring to storm, noise) zaxa 'won' zixa 'made win* 'ayaf 'got tired' 'iyef 'caused to get tired' (+tr) rina zaxta (Bl) bapras hariSon. Rina won 3p.s.f. in-the-price the-first 'Rina won the first price' hatluS haze mezake (B3) et ba'alo beoto. the-ticket the-this makes-win/ GM owner-its in-car makes entitled 3p.s.m. 'Due to this ticket its owner wins a car'.

There is a fairly large group of roots that are intransitive in the PAAL pattern and transitive in the PIEL·. Nevertheless, the derivation of PIEL as the transitive of PAAL cannot be called productive in NH because no new roots are added to this paradigm, nor is there psycholinguistic evidence for its productivity. (Empirical tests (Bolozky 1983) demonstrated that new transitive verbs enter the system in PIEL·, and to some extent HIFIL·, and new intransitive verbs in HITPAEL.) Productive valency increase is the derivation of causative HIFIL· from PAAL. The next section will be about this derivation.

74

4.5.1 Causativity - a working definition Before going into the valency increase operation for deriving the HIFIL, the term 'causative' has to be clarified because, like 'transitive', it is used in various ways. The term causativity is usually used to refer to a State of Affairs in which a causer makes the causee perform an action, undergo a process (or experience) or be in a state. The causer may be a human Agent, as in (26) or a Force, as in (27). The causee too may be human, as in (26), or a non-human entity as in (27). (26)

(27)

bezman hatironut hasamal haya mac'id (B5) otanu in-time-of the-training the-sergeant used-to-make-march us kol boker sa'a. every morning hour 'During the training the sergeant used to make us march every morning for an hour'. haruah hir'ida (B5) et ha'alim. the-wind made-tremble GM the-leaves 'The wind made the leaves tremble'.

The definition of the causative construction may focus on the situation, (see Comrie (1981)), on the aspects of logic and truth value, (see Shibatani (1976)), on syntactic features and syntactic derivation, (see Cole (1976b and Herman (1979)), or on the syntactic process and the semantic changes of the participants, (see Saad & Bolozky (1980, 1984)). Dik (1980:ch. 3) describes causativity as a process of predicate formation by which the predicate-frame of an input predicate undergoes an increase in the number of arguments, and a change in meaning. A causative predicate has the following meaning: (28)

A causer xO makes a causee xl perform the act or be in the state designated by the predicate.

Oik (ibid) suggests that the output predicate-frame gets an additional argument, (xO), which has the semantic function Causer, in all causative constructions. Vet (lecture at the University of Amsterdam, 3.3.86) put forward an alternative FG approach to causativity, suggesting that the arguments of a causative verb should be assigned more precise semantic functions than causer and causee. He also distinguishes between three types of causative constructions, which in Dutch have three different underlying representations.8 There is no reason, however, to adopt this distinction for MH. MB has two types of productive morphological causatives with the HIFIL, their difference being determined by the semantic function of the causee. But this does not amount to a different construction which would justify different derivational rules. The definition of causative constructions used here is as follows: (29)

a causative construction is a construction which designates a State of Affairs S in which a causer (xO) creates a State of Affairs S' in which a causee (xl) performs an act, undergoes a process or is in the state designated by the input predicate.

Causativity in Hebrew and Arabic has been described by Saad and Bolozky (1980, 1984) and Hentcher (1983).

75

Noutaouakil (lecture at the University of Amsterdam, September 1986) shows that in Arabic there are three types of constructions for expressing causativity: 1. lexical: causativity is expressed by means of a predicate which is itself causative, such as 'kill'; 2. periphrastic: causativity is expressed by means of a verb of causation like 'cause to/ make to + another predicate; 3. morphological: the causativity is expressed by the morphological pattern of the verb. In standard Arabic these patterns are the 'af'ala and the fa'la. Hebrew has the same system for expressing causativity. The periphrastic causatives are formed with the verbs garam 'caused' or 'asa Se 'made that'; garaa is more commonly used. The morphological causatives can be formed with the HIFIL and PIEL patterns. 9 Of the two, only the HIFIL is productive. The HIFIL· cannot be used as the causative of every root, but it meets both demands for being considered productive: speakers recognize and use it as such, and new causative verbs are formed with it. (Although, this formation is not automatic, and to some extent the verbs expressing causative with HIFIL· have to be learned as such.) In colloquial Hebrew one can find the neologism (30)

labac 'stress'

—>

hilbic 'cause stress, put under stress'

which I heard used for example in: (31)

aval bexol zot divrey hapsixologit *al hayeled but still words-of the-psychologist on the-child hilhicu (B5) et isti meod. made-nervous GM wife-my very 'But still what the psychologist said about the child made my wife very nervous*.

The picture is made slightly more complex by the fact that there are also verbs in the HIFIL· which are no longer causative, such as (32)

hod 'take out' (y.c.a. 'exit' + HIFIL·) bimci 'invent' (m.c.a. 'find' + HIFIL·)

Such verbs in the HIFIL· are lexicalized causatives, which at an earlier stage of the language may have been derived by a causative formation rule, but which in the meantime have acquired a meaning that is not directly produced by the root and causation. 10 The derivational processes of these verbs are, therefore, considered regularities and are listed in the lexicon as idiosyncracies. The following picture emerges: (33)

lexicon derived

basic + regularities causative

HIFIL·




HIFIL·

The rest of this section will be about the productive causatives based on the HIFIL·. The following sentences illustrate several HIFIL· causative constructions:

76

(34)

yosi heexil (B5) et dan et ba'uga. Yosi made-eat 3p.s.m. GN Dan GM the-cake 'Yosi made Dan eat the cake* haruah hir'ida (B5) et ba'alim the-wind made-tremble 3p.s.£. GM the-leaves 'The wind made the leaves tremble' hasapar hoäiv (B5) et hayeled "al hakise. the-hairdresser made-sit 3p.s.m. GM the-child on the-chair 'The hairdresser made the child sit on the chair* hara'aS hirgiz (B5) oti. the-noise made-angry 3p.s.m. me 'The noise annoyed me*.

(35) (36) (37)

The underlying predication of these constructions is: (38)

pred V-B5 (xO)Ag/Fo(causer) (xl)(causee) [(x2)Go/Proc/Phen] ( t ] « optional) Note the following points: 1. (xO) is the causer. In MH it is always in the 'nominative case', (i.e. without any prefix, particle or marker) and its semantic function can be Ag or Fo. 2. (xl) is the causee. Its semantic function has been left unspecified for the time being, and it will be dealt with in detail in section 4.5.2. The causee in HIFIL causative constructions is always preceded by the particle et or Je. Traditionally these are seen as the accusative and dative markers respectively.11 In order to be able to formulate the rules deriving the causative constructions with HIFIL, it is necessary to look now first at its predicate-scheme, in particlular at the semantic function of the causee. 4.5.2 The semantic function of the caueee Consider again several causative constructions: (39)a b c (40)a b c

reuven heexil (B5) et hayeled et hatapuah. Reuven made-eat 3p.s.m. GM the-child GM the-apple 'Reuven made the child eat the apple', haem hilbisa (B5) et hayeled et hameil. the-mother made-dress 3p.s.f. GM the-child GM the-coat 'The mother made the child wear the coat', hasofet hisbi'a (B5) et ha'ed et hasvu'a. the-judge made-swear 3p.s.m. GM the-witness GM the-oath 'The judge made the witness take the oath'. hizkarti (B5) leyosi et havtahato. made-remember-lp.s. to-Yosi GM promise-his reminded Yosi about his promise', hasar hodi'a (B5) la-'itonaim et hahadaSot. the-minister made-known 3p.s.m. to-the-reporters GM the-news "The minister announced the news to the reporters', dan hismi'a (B5) le-yosi et taklito hahadaS. Dan made-bear 3p.s.m. to-yosi GM record-his the-new 'Dan made Josi listen to his new record'.

In (39a-c) the causee is preceded by the particle et, and in (40 a-c) it is preceded by the prefix Je. Regarding these two HIFIL· causatives, with et + causee and Je + causee, the following questions are of interest:

77

1. what kind of particle/ prefix are et and le ? 2. when do they occur in causative constructions? 3. what is the relation of these particles/prefixes to the semantic function of the causee? These three questions are intertwined to such an extent that answering one automatically implies partially answering the others too. To avoid unnecessary repetition the questions shall not be dealt with separately but in conjunction with each other. In most cases et is the particle which precedes the causee. Usually, there is only one constituent with et. Sometimes a second et constituent occurs, as in sentences (41)-(42) below. This second et constituent is optional. It is possible to obtain perfectly good HIFIL causative constructions using only one et causee constituent. When there are two Goals, several syntactic differences are involved. Cole (1976b) notes the following ones: (I present only the essence of Cole's observations, and transfer his terminology and theoretical frame from Transformational Generative Grammar to Functional Grammar.) The differences between the Go causee, referred to as Gol, vs. the Go in the 'caused State of Affairs', referred to as Go2, are the following: (i) Only Gol can be cliticized. In the sentences below (a) is the full uncliticized construction, in (b) Gol is cliticized and in (c) Go2: (41)a b c (42)a b c

hirkadeti (B5) et hatalmidim et harikud haze. made-dance-lp.s. CM the-students GN the dance the-this Ί made the students dance this dance*, hirkadetim (B5) et harikud haze. made-them-dance-lp.s. GN the-dance the-this 'I made them dance this dance'. «hirkadetim (B5) et hatalmidim made-dance-them-lp.s.-the-dances GN the-students hisb'ati (B5) et ha'ed et hasvu'a hazot. made-swear-lp.s. GM the-witness GN the-oath the-this Ί made the witness swear this oath', hiSba'ativ (B5) et haSvu'a hazot. made-him-swear-lp.s. GN the-oath the-this I made him swear this oath*, «hiSba'ativ (B5) et ha'ed made-swear-it-lp.s. GH the-witness

(ii) Causative constructions are rare in the 'passive', but when they do occur, only Gol is accessible to assignment of Subj function, never Go2. (41)d e (42)d e

hatalmidim hurkedu (B6) et harikud hahadaS. the-students were-made-to-dance-3p.pl. GN the-dance the-new 'The students were made to dance the new dance', «harikud hahadas hurkad (B6) et hatalmidim. the-dance the-new was-made-to-dance-3p.s.m. GN the-students ha'ed huSb'a (B6) et haSvu'a hazot. the-witness was-made-to-swear-3p.s.m. GN the-oath the-this 'The witness was made to swear this oath', «hasvu'a hazot husbe'a (B6) et ha'ed. the-oath the-this was-sworn-3p.s.f. GN the-witness

Since causative constructions are all inherently transitive, and et is the GN, it could be seen as having the same function in causative constructions as well. The underlying predication of such a causative construction would be as follows:

78

(43)

C.C.C. V-B5 (xO)Ag/Fo(causer) (xl)Go(causee) [(x2)Go] ([ ] indicate optionality)

Causative constructions, as simple active-transitive ones, cannot always undergo valency reduction and become pseudo-passive. Furthermore, even when valency reduction is possible, there is a strong tendency to leave the Ag(Causer) argument unspecified. The possible 'passive' equivalents of sentences (39a-c) are: (39)a* (39)b' (39)c*

?dan huaxal (B6) 'al yedei reuven Dan was-made-to-eat-3p.s.m. by Reuben Phayeled hulbas (B6) 'al yedei haem 1Z the-child was-dressed-3p.s.m. by the-mother ?ha'ed huäb'a (B6) 'al yedei haSofet the-witness was-made-to-take-oath-3p.s.m. by the-judge

According to Cole (1976b) the verbs can take two arguments preceded by et are the following: (44)

heexil 'caused to eat, f e d ' , hifSit 'made undress' hirkid 'caused to dance', hiSbi'a 'caused to swear', biSka 'caused to drink' and hilbiS 'caused to dress'. (not mentioned by Cole (ibid) is he'evir 'made cross/pass')

(In fact in colloquial Hebrew also heexil le ( ' f e d 1 and hifSit le 'made undress' are used.) Note that the two et constituents are syntactically not identical. Only (xl) is Go of the HIFIL verb, (x2) is the Go in the 'caused State of Affairs'. At first sight it seems that assigning Go function to the causee preceded by et poses a problem in constructions like (45)

dan hirgiz (B5) Dan made-angry 3p.s.m. 'Dan annoyed Yosi'.

et yosi. GM Yosi

Does the fact that hirgiz 1'annoy' is an emotive verb not imply that the semantic function of 'Yosi is Experiencer and not Go despite the particle et? Here I want to introduce a distinction between the state of affairs designated by the whole causative construction, and the 'caused state of affairs'. The whole 'matrix' construction designating a certain state of affairs called S contains within it a reference to another state of affairs, S ' , which is the 'caused state of affairs'. Hence (45) is S, and within S there is reference to another state of affairs S' - the one that is caused. The 'caused state of affairs' in (45) is (46)

yosi hitragez (B7) Yosi got-angry 1 'Yosi got angry

and here Yosi is indeed Experiencer (Note that it also occurs with the intransitive HITPAEL pattern). But sentence (46) refers to a different State of Affairs, featuring Dan as the Ag who has committed an annoying act; Yosi is the Go of that act. Therefore, I think it is justified to conclude that the particle et has the function of GM in causative constructions, which in fact are a subgroup of the group containing all the transitive constructions. This

79

holds for the majority of the causative constructions based on the HIFIL. At first sight the prefix le would appear more problematic as it can mark several semantic functions. These are: Recipient, Beneficiary, Experiencer and Direction. In causative constructions Je occurs with a fairly limited number of verbs. The list of Je-verbs in (47) below is an expansion of a list given by Cole (1976b). Note that bilbiS 'cause to dress' can take both particles. (Cole also mentions hisJcir 'rent' which in my opinion is a lexicalized causative and not derived by a synchronic rule from the PAAL.) (47)

bodi'a 'made known, announce', bixtiv 'made write, dictate', biSai'a 'made hear', 1hikna 'made aquire', hizkir 'reminded', bera 'showed , boriS 'made inherit', biohil 'bequested, biskir 'rented', hilbiS 'made dress, made wear' hin'im 'made pleasant', bixiv 'caused pain'

It is difficult to give any further characterization of these verbs as a group. Semantically they are not homogeneous; 55% of them, namely bodi'a 'made known', biSai'a 'made heard', bera 'showed', bizkir 'reminded 1 , bin 'im 'made pleasant' and hikna 'made aquire' (in the idiom hikna yeda 'made acquire knowledge') are the causative form of verbs of perception and cognition, but the rest are not. The majority of the verbs of perception and cognition do not have Ag as their first argument but Experiencer (Exp) or Processed (Proc). For example (48) (49) (50)

et hataklit haze kvar sam'ati. (Bl) (Exp) GM the-record the-this already heard-lp.s. 'I've already heard this record'. Saul (Proc) yaras (Bl) et haoto haze medodo. Saul innerited-3ps.m. GM the-car the-this from-his-uncle 'Saul inherited this car from his uncle* na'am li(Exp) lismo'a ( i n f ) et kol chokam sei was-pleasant to-me to-hear GM voice laughter of hayeladim the-children Ί liked hearing the sound of the children's laughter'

Exceptions where xl is Ag are bixtiv 'made write', bikna 'made acquire' and biskir 'rented'. Causativization is in fact a process of transitivization. If the verb was intransitive, an extra argument is added: the causer (which creates the State of Affairs designated by the predicate in the PAAL.) The original predication (with PAAL) becomes the 'caused State of A f f a i r s ' . With predicates taking et + causee the argument which was Ag in the original State of Affairs becomes either Ag(causee) or Go in the causative State of Affairs, depending on the valency of the HIFIL.The first argument, (xO) is always Causee. When the causative verb is trivalent, (xl) is Ag(causee) and (x2) is Go; when the causative verb is bivalent (xl) is Go. With predicates taking le + causee the Ag in the original State of Affairs becomes Exp or Proc in the 'caused State of A f f a i r s ' ; this, however, holds only for the constructions with verbs of perception and cognition or emotive verbs; otherwise, the (xl) remains Ag(causee) or Go (here too depending on the valency of the causative verb) as with k.t.b. 'write', y.r.S. 'inherit' and others. There are, however, emotive verbs which take the particle et: birgiz 'made angry' and he'eciv 'made sad'. This raises an interesting problem: is the semantic function of (x2) determined here by the particle, in

80 which case it is Go; or is it determined on the basis of consistency with the group of verbs of perception and cognition, in which case it is Experiencer? Causatives with the FIEL also include emotive verbs. For example: (51)

yosi 'icben (B3) et Yosi made-angry 3p.s.m. GM 'Yosi made everyone angry 1 .

kulam. everyone

In Junger (1985a) et is analysed in such PIEL causatives as Goal Marker. I think that this should be maintained for the HIFIL causatives with emotive verbs, rather than classify them with the verbs of perception and cognition. There is also an independent reason why these two emotive verbs differ from the verbs of perception and cognition. Whereas in causative constructions verbs of perception and cognition have an obligatory (x2)Go, (i.e. Go of the causee) the emotive verbs never do. Note that they do have an (xl)Go, which is the Go in the causative construction (i.e. the State of Affairs designated by the predication with HIFIL·) and Exp in 'caused State of A f f a i r s ' . (52)

hahadaSot 'al moto Sei hanasi he'ecivu (B5) et the-news on death-of of the-president made-sad 3p.pl GM kulanu. us-all 'The news of the president's death made all of us sad'.

The same holds for birgiz 'made angry'. Therefore, Ae'ecirand hirgiz have the following predicate frame:

(53)

pred V-B5 (xO)Ag(causer) (xl)Go(causee)

Let us return to the particles et and Je and the HIFIL constructions in which they occur. The two types of HIFIL-causatives differ in yet another respect: in the constructions with a double et the (x2) is optional, but in those with Je preceding (x2), the (x2) is obligatory. In (39a-c) (x2) can be left unspecified, but in (42a-c) (x2) must be specified. The two types of HIFIL causatives can be summed up as follows: (xO)Ag/Fo (xl)Go(Ag causee) [(x2)Go]

(54)

C.C.C. V-B5 (xO)Ag/Fo (xl)Exp(Exp causee) (x2)Go/Phen

The semantic function of the xl argument in the 'caused state of affairs' is given in the brackets following the semantic function of the arguments in the causative construction.

4.5.3 Derivation Dik (1980:78-79) suggested the following predicate formation rule for Dutch causatives with laten:

81 (55)

CAUSATIVE PREDICATE FORMATION

input

: any predicate-frame Pred V (xl)..(xn) where (xl) does not have 0 function (that is, the predicate frame does not designate a State) output : laten-Pred V (xO)Ag (xl).. (xn) meaning: "xO causes the state of affairs designated by the input predicate frame to take place". Let us see if this analysis is also applicable to the HIFIL causatives. The first condition which will have to be met is that of productivity. In the above mentioned frequency list of MR compiled by Balgur (1968) there are 276 roots which can occur in the HIFIL pattern. Of these, 149 are causative (108 productive and 41 lexicalized), 31 are active transitive and 2 are inchoative. Despite its considerable frequency there are also verbs which do not occur at all in the HIFIL·, although there is no morphological or semantic reason for this. For example (56)

Savar 'broke' Saaf 'wished 1 oa t an 'gave'

— *biSbir 'caused to break' — *hiSif 'caused to wish' — * hint in 'caused to give*

The verbs in (56) and others like them form the causative analytically, i.e. by means of an auxiliary causative predicate. However, the group of verbs which do not occur in HIFIL is smaller than those which do, and they seem to form an accidental gap. I could not detect any semantic or syntactic common denominator linking them. The HIFIL· also has psychological reality as a causative pattern. Bolozky (1978) reports four productivity tests, which indicate that there is a preference of 60% to form causatives with HIFIL· (the other 40% going to FIEL). Waiden (1982) established that the HIFIL is used productively to form causatives also by children. For example, children accept the nonstandard heehiv 'caused to love', lehavkot 'cause to cry', hilhic 'put under pressure' is by now an accepted expression in colloquial Hebrew. This would seem to justify the conclusion that the HIFIL· pattern meets the conditions for productivity, and can be derived by means of a predicate formation rule. Such a rule would be (57)

CAUSATIVE HIFIL FORMATION input : C.C.C. V-B1 ( x l ) . . ( x n )

output: a. C.C.C. V-B5 (xO)Ag (causer) (xl)Go (Ag causee) [(x2)Go2] b. C.C.C. V-B5 (xO)Ag/Fo(causer) (xl)Exp (Exp causee) (x2)Go meaning: "xO makes the state of affairs designated by the input predicate apply to xl in the output". Note the following points: 1. Rule (57) is more detailed than rule (55)in that it gives a more specified constraint on the form of the input predicate- it has to be in the PAAL or be a root which can occur in the PAAL. This is necessary because there are roots who have another bioyan, the PIEL for example, as their basic active pattern. These do not form their causative with the HIFIL·, as illustrated below: (58)

Silem (B3) 'paid' dyer (B3) 'drew' siper (B3) 'told'

— *hiSlim 'caused to pay' — *hicyir 'caused to draw'1 — *hispir 'caused to tell

82

2. This rule does not exclude the input of verbs which like the ones listed in (56) simply do not occur in the HIFIL at all. I think that the lexicon makes it clear that they form an accidental gap. For the rest, all these roots whose basic active pattern is the PAAL and which can occur in the HIFIL·, can be input to rule (57); the output will be the root in the HIFIL· pattern, expressing causativity. 4.6 Valency reduction 4.6.1 Valency reduction vs. Subject assignment to Go This section will treat the biayaoim HIFAL·, PUAL and HUFAL, which as we shall see are derived from the PAAL, PIEL and HIFIL· respectively. In principle, FG offers two possibilities for deriving them: by assigning the syntactic function Subject to Go, or by valency reduction. There are two syntactic functions in FG: Subject and Object, both of them expressing the perspective from which a certain state of affairs is regarded. Thus the State of Affairs designated by the predication (59)

Mary gave this book to Jane

can be viewed from two perspectives: Mary's and the book's. These are captured in the following underlying predications: (60)a b

give V (dlxl :Mary(xl))AgSubj (dlnx2 :book(x2))Go (dlx3 :Jane(x3))Rec give V (dlxl :Mary(xl)Subj (dlpx2) :book(x2))GoSubj (dlx3 :Jane(x3))Rec

Which are realized by the active and passive voice respectively. (61)a b

Mary gave this book to Jane. This book was given to Jane by Mary.

In other words, assignment of Subj function to construction. The possibilities of assigning Subj arguments depends on their semantic function, and language. Subject designates primary perspective. distinguish also between 'secondary perspective'. (59) this means the following opposition: (62)a b

Go results in a passive function to the various varies from language to Some languages In the case of sentence

Mary gave this book to Jane, Mary gave Jane this book.

The difference in the secondary perspective is in having assigned Obj function to the Go or the Rec respectively. (63)a b

give V (dlxl :Maryxl))AgSubj (dlp(x2) :book(x2))GoObj (dlx3 :Jane(x3))Rec give V (dlxl:Mary(xD)AgSubj (dlpx2 :book(x2))Go (dlx3 :Jane(x3))RecObj

In this case the difference in Obj assignment is realized by a different word order. MH does not make a distinction of secondary perspective, i.e. has no Obj assignment. It could, however, be claimed that the binyanim NIFAL, PUAL and HUFAL express a difference in primary perspective, i.e. thay express passive voice. In such a case they would be derived from an underlying predication where the verb would be in the

83

active binyan (i.e. PAAL, FIEL or HIFIL) and the Go argument would be assigned Subj function. For example (64)a

b

'acar V-B1 (dlxl: mistara(xl))AgSubj (dlx2 :poSe'a(x2))Go

'acar V-B1 (dlxl: miStara(xl))Ag (dlx2 :poSe'a(x2))GoSubj

These underlying predications would be realized as follows: (65)a b

hamiStara 'acra et hapofie'a the-police arrested 3p.s.f. CM the-criminal 'The police arrested the criminal' hapose'a ne'ecar 'al yedei hamiStara the-criminal was-arrested 3p.s.m. by the-police 'The criminal was arrested by the police*.

This analysis is rejected on basis of the MH data, as will be shown in section 4.6.2 . The other possibility of deriving the NIFAL, PUAL and HUFAL patterns is by valency reduction. This is the approach addopted in this study. The operation of valency reduction in FG is described in section 4.6.2 below, and its application to MH in section 4.6.3 . 4.6.2 Valency reduction in FG In section 4.5 we saw the operation of valency increase and its application to MH. The other productive quantitative valency change in the binyania system is valency reduction. Valency reduction, or detransitivization, is described in Dik (1983) by the following general rule: (66)

DETRAMSITIVIZING PREDICATE FORMATION

input : pred V (xl) (x2) output : pred V-R (xl) meaning: "the relation expressed by pred V applies to (xl)" 'R'stands for some sort of marking of the derived predicate (which sometimes has no explicit form). As we shall see, in the case of Hebrew it stands for the possible binyan of the output predicate. Note that this rule can be applied to both the first and the second argument. When the first argument is reduced the output predicate is interpreted as pseudo-passive, and when the second argument is reduced the output predicate can be interpreted as reflexive, reciprocal or de-actualized (generic, habitual). Both these types of valency reduction exist in Hebrew, and will be described in the rest of this section. 4.6.2.1 First argument reduction: the NIFAL, PUAL and HUFAL pattern·

Consider the following pairs of sentences, which illustrate the use of the NIFAL, PUAL· and HUFAL: (67)a b

hamiStara 'acra (Bl) beezor tel-aviv 6 haSudim. the-police arrested 3p.s.f. in-area-of Tel Aviv 6 suspects 'The police arrested in the area of Tel Aviv 6 suspects', *beezor tel aviv ne'ecru (B2) 6 haSudi· in-area-of Tel Aviv were-arrested 3p.pl.rn. 6 suspects 'al yedei hamistara. by the- police

84

c (68)a

b

c

(69)a

b

c

beezor tel aviv ne'ecru (B2) 6 haSudim. in-area-of Tel Aviv were-arrested 3p.pl.m. 6 suspects 'Six suspects were arrested in the Tel Aviv area 1 . hakneset iära (B3) et resimat hasarim kfi the-parliament approved 3p.s.f. GM list-of ministers as sesamir hevi ota. that-Samir brought her 'The parliament approved the list of ministers presented by Shamir' *resimat hasarim sesamir hevi list-of the-ministers that-Shamir presented 3p.s.m. usra (B4) 'al yedei hakneset. was-approved 3p.s.f by the-parliament resimat hasarim kfi Sesamir hevi ota list-of the-ministers as that-Shamir presented her usra (B4) was-approved 3p.s.f. 'The list of ministers presented by Shamir was approved'. hamisrad yagis (B5) et hatoxnit the-dept. will-present 3p.s.m. GM the-plan le'iyuno Sei sar hoacar. to-consideration of minister of economy 'The department will present the plan for the consideration of the minister of economy', *hatoxnit tugas (B6) le'iyuno Sei the-plan will-be-presented 3p.s.f. to-consideration of sar-haocar 'al yedei hamisrad minister of economy by the office hatoxnit tugas (B6) le'iyuno Sei the-plan will-be-presented 3p.s.f. to-consideration of sar haocar minister-of economy 'The plan will be presented for the consideration of the minister of economy*

These examples show that in constructions with the so called 'passive' patterns NIFAL (B2), PUAL (B4) and HUFAL (B6) there is normally no Ag-phrase. The addition of an Ag to such constructions is rare and produces very strange sentences. This characteristic of the 'passive' constructions in Hebrew has also been discussed in Berman (1979a). However, the absence of Ag seems to be not limited only to Hebrew, but a more general characteristic of these constructions. This has led Frajzyngier (1982), for example, to connect passives to impersonal constructions in general. (He looked at data on passives in Polish, Turkish, Latin, Italian, Arabic, German and Dutch.) Examining the corpus compiled for this study, I found that an Ag phrase is only allowed when it is required for a specific pragmatic reason. An Agent-phrase can occur in the following cases: 1. when the identity of the Agent is important, and it is not clear from the context: (70)

cav hagbala minhali aSer huca (B6) order limitation administrative which was-issued 3p.s.m. negdi al yedei siltonot habitation... against-me by authorities-of security... 'The administrative limitation order which1 was issued against me by the security authorities...

85

2. when the Ag is a proper name: (71)

bevrat sleyd meyuceget (B4) baarec 'al yedei bevrat company Sleyd represented 3p.s.f in-Israel by company ardo Ardo 'The Sleyd Company is represented in Israel by the Ardo Company'.

3. when the Ag is emphasized, usually as a contrastive emphasis. (72)

hamehirim nikbe'u (B2) lo 'al yedei hevrat the-prices were-fixed 3p.pl. not by company-of habasmal ela 'al yedei misrad hapnim. electricity but by ministry-of internal affairs 'The prices were fixed not by the Electricity Company but by the Ministry for Internal A f f a i r s ' .

In these sentences, as in all sentences featuring an Ag phrase, the Ag- phrase can be omitted. In fact, in a sample of 47 sentences with these patterns only 7, i.e. 14.6%, included an 'al yedei ( ' b y ' ) phrase. This frequency suggests that the Ag-phrases in these sentences are satellites of Ag, Force, Cause or Reason rather than arguments. Further reasons why this Ag-phrase should be seen as a satellite rather than as an argument are presented in Junger (1986a). Briefly these are: 1. The output predicate of such a detransitivization rule can designate a different type of State of Affairs than the input predicate. Whereas the input predicate is sometimes ^Control, its detransitivized equivalent is -Control. (A predication is +Control when it includes a constituent which can make it take place, or prevent it from taking place. Otherwise it is -Control.) This can be demonstrated by adding a phrase that expresses volition, like bexavana On purpose'. (73)a b

yosi Savar (B1) et hakos bexavana. Yosi broke 3p.s.m. GN the-glass on-purpose 'Yosi broke the glass on purpose 1 , hakos niSbera (B2) (*bexavana) the-glass broke 3p.s.f. (* on purpose)

It is, however, possible to add an expression of 'autonomy', suggesting that the event happened by itself, without the agency of someone: (74)c

hakos niSbera (B2) me'acma. the-glass broke 3p.s.f. by-herself 'The glass broke by itself.

2. An 'al yedei phrase can also occur in a sentence with an active verb form. (75)

hapoSe'a hitabed (B7) bakele yomayim lifney the-criminal killed-himself in-prison two-days before hamiSpat 'al yedei bli'at kadur ra'al. the-trial by means of swallowing-of pill-of poison 'The criminal committed suicide in the prison two days before the trial by swallowing a poison pill.'

86

3. In some cases where an 'al yedei phrase is expected, there is a satellite of Location or Source instead. (76)

hanetunim 'ubdu (B4) bamahSev. the-data Here-processed 3p.pl. in-the-computer 'The data were processed in the computer.' mimisrad dover cabal nimsar (B2) sekohotenu from-office-of spokesman army was-announced that-forces-our Savu beSalom.

(77)

returned in-peace 'From the army spokesman's office it was announced that our troops returned in peace.' In conclusion, the constructions with the NIFAL, PUAL and HUFAL patterns described in this section do not result from Subj assignment to Go. Rather, they are constructions where the first argument (Ag) has been reduced. The result is a construction which resembles a passive, but is not really one. (For more details on passivization in FG, see Dik 1978). Derivation If it is accepted that the bioyaoia NIFAL·, PUAL and HUFAL in the constructions discussed so far are not passive but detransitivized, they will be derived by rule (70) (suggested in Dik 1983a): (78)

PSEUDO- PASSIVE FORMATION input : pred V (xl)Ag (x2)Go output : pred V (x2)Proc

The rules for MB can be formulated as follows: (79)a

NIFAL· FORMATION

input : output : b

PUAL FORMATION

input : output : c

C.C.C. V Bl (xl)Ag (x2)Go C.C.C.V B2 (xl)Proc C.C.C.V B3 (xl)Ag (X2)Go C.C.C.V B4 (xl)Proc

HUFAL· FORMATION

input : output :

C.C.C.V B5 (xl)Ag (x2)Go C.C.C.V B6 (xl)Proc

Rules (79a-c) hold for the productive pairs, where the NIFAL, PUAL· and HUFAL· occur in pseudo-passive constructions. The picture, however, is somewhat more complex, because there are also roots where one transitive pattern undergoes detransitivization and yields two different patterns. For example (80)

ilec (B3) 'forced' hizkir (B5) 'remembered'

ulac (B4) 'was forced* huzkar (B6) 'was reminded'

hexsil (B5) 'made fail' -

huxsal (B6) 'was made to fail'

neelac (B2) 'was forced' nizkar (B2) 'was reminded/ remembered' nixsal (B2) 13 'failed'

87

Since the root a.I.e. does not occur at all in the PAAL, the NIFAL must in this case, be derived from the FIEL too, like the PUAL. Such cases with a second detransitivized pattern are not frequent. These patterns are discussed among others in Stern (1979), who points out that there is a semantic difference between the PUAL and HUFAL on the one hand and the NIFAL· on the other. He calls the former two 'event passive' {Vorgangspassiv) and the latter 'state passive 1 (Zustandspassiv). In terms of FG this means that the PUAL and HUFAL refer to a +dynamic State of Affairs, and the constructions with the NIFAL refer to a -dynamic State of Affairs. A -»-dynamic State of Affairs is defined as involving change. Looking at constructions with PUAL and HUFAL· vs. those with NIFAL· I could not find any such difference. However, I did find that the addition of an 'al yedei phrase is more acceptable with the PUAL· and HUFAL pseudo passives than with the NIFAL ones. This is illustrated below: (81)a

b

rina huxSela (B6) bemivhan hanehiga bexavana Rina was-made-to-fail in-exam-of driving on-purpose 'al yedei yosi, haboben hahadaS. by Yosi, the-examiner the-new 'Rina was made to fail the driving exam on purpose by Yosi, the new examiner' *rina nixSela (B2) bemivhan henhiga bexavana Rina failed 3p.s.f. in-exam-of driving on-purpose 'al yedei yosi, haboben hahadaS. by Yosi, the-exaainer the-new

It would seem that the NIFAL· construction does indeed refer to a State or result, rather than to an Action or event. The difference between a State and an Action type State of Affairs regarding the first argument is in the parameter Control. In a State the first argument (as the whole predication) has the feature [-Control], whereas in an Action State of Affairs the first argument is [^Control]. Indeed (82) below, which is also a [-Control] State of Affair is with the NIFAL. (82)

rina nixSela (B2) bemivhan hanehiga Selo Rina failed 3p.s.f. in-exam of-driving that-not beaSmata ela biglal Se... her-fault but because that... 'Rina failed the driving exam through no fault of her own but because...'

The NIFAL· expresses State mostly with a verb that also occurs in the HIFIL (causative or transitive). In fact rules (80a-c) do not account for these cases because they are not productive. The NIFAL· in such constructions will be dealt with in chapter 5 section 5.5.1.1 . 4.6.3 Second argument reduction According to Dik (1985) and Dik & Gvozdanovic (1981) reduction of the second argument may yield reflexive, reciprocal and intransitive interpretations. All of these can be found in Modern Hebrew and will be described in the rest of this section.

88

Reflexivity Reflexivity can be expressed in Modern Hebrew in two ways: morphologically, i.e. by means of the form of the verb, or by means of reflexive pronouns. 14 The former is illustrated in (83) and the latter in (84) below: (83)a b (84)a

yosi mitgaleah (B7) kol yom. Yosi shave-himself every day 'Yosi shaves himself every day', hu histaken (B7) banesi'a hazot. he endangered-himself in-trip the-this 'He endangered himself in this trip'. hu raa (Bl) et 'acmo bezgugit haoto. he saw 3p.s.m. GM himself in-window-of the-car 'He saw himself in the car window'.

The HITPAEL-reflexive can almost always be replaced by a pronominal reflexive, but the opposite does not hold. (83)c d

?yosi megaleah (B3) et 'acmo kol boker19 Yosi shaves 3p.s.m. GM himself every morning hu siken (B3) et 'acmo banesi'a hazot. he endangered 3p.s.m. GM himself in-trip the-this 'He endangered himself in this trip'.

but

(84)b

*hu hitraa (B7) bezgugit haoto. (if hitraa = reciprocal) he saw 3p.s.m. ίη-window of-the-car

Strictly speaking sentence (83c) is not correct, as use of the pronominal reflexive instead of the HITPAEL is one of the features of sub-standard (but definitely existing) Hebrew. Reflexivity is expressed pronominally rather than by the HITPAEL if: 16 a. the HITPAEL of the root in question has another function (reciprocal, intransitive or inchoative) b. the root in question does not occur in the HITPAEL at all, due to gaps in the system. The complete Hebrew Dictionary (Even-Shoshan 1967) lists 269 verbs in the HITPAEL form. Of these, only 32 have a reflexive primary meaning. The majority of the verbs, 165, are intransitive. Nevertheless, Hebrew speakers regard the HITPAEL as a productive way of expressing reflexivity. It also seems to create new expressions such as (85)

hatix 'good looking' yafe 'beautiful'

—> —>

hithatex 'made himself look good' hityape 'made himself beautiful'

In these cases, however, the reflexive verbs, derived from nouns or adjectives, are not cases of detransitivization (since the root does not occur in any other verbal pattern). The productivity of the HITPAEL as reflexive form seems to be decreasing. A study on the use of the bioyanim by children of the ages 3-6 (Berman, 1982) shows that the HITPAEL is used primarily as an active intransitive form, and less frequently to express iterativeness, reciprocity and inchoativeness. The tendency in colloquial Hebrew to replace the HITPAEL by the pronominal reflexive among grown-ups too has been noted earlier (Berman, 1978). Bolozky (1978) conducted a survey of four productivity tests of the binyaniot. His results showed a

89

preference of almost 100% to use the HITPAEL as inchoative or as reflexive. My findings in Even-Shoshan (1967) as well as in my own corpus show that the HITPAEL is primarily intransitive. (In 75 sentences with 93 verbs in the HITPAEL, 83 were intransitive and 4 were reflexive.) The difference may be due to contextual differences between the constructions used in my corpus and those used by Bolozky to elicit the verb forms. In any case, it seems justified to regard the reflexive HITPAEL as productive and to derive it by a predicate formation rule of 2nd argument reduction. This rule is as follows: (86)

REFLEXIVE HITPAEL FORMATION

input : output : meaning :

C.C.C. V B1,B3 (xl)Ag (x2)Go C.C.C. V B7 (xl)Ag the action designated by the verbal predicate is applied by the Ag (xl) to himself/herself

Note that the input predicate to this rule can be a verb either in the PAAL or in the PIEL form. In addition, the semantic function of the (xl) argument in the output does not change as it did in the first argument reduction. Reciprocal Reciprocity, like reflexivity, can be expressed both by the HITPAEL pattern, and pronominally. The HITPAEL· pattern in reciprocal constructions is regarded as detransitivized. Consider the following examples, in (a) with the HITPAEL· and in (b) with a pronominal expression: (87)a b

(88)a

b

bemeSex hodasim yosi vedan hitrau (B7) kol yom ... during months Yosi and Dan saw -rec. every day ... 'For months Yosi and Dan saw each other every day—' bemeSex hodaSim yosi vedan hayu roim(Bl) / rau(Bl) ze et ze during months Yosi and Dan was see /saw this GM this kol yom every day 'For months Yosi and Dan used to see/ saw each other every day...' bemeSex 10 sanim hu hitkatev (B7) 'im ahiv bemoskva during 10 years he wrote-rec with brother-his in-Moscow 'ad Sesof sof nifgeSu (B7). until finally met-3p.pl.(rec) 'For 10 years he corresponded with his brother in Moscow until they finally met.' bemeaex 10 Sanim hu veahiv bemoskva during 10 years he and-brother-his in-Moscow hayu kotvim(Bl) / katvu (Bl) ze leze 'ad Sesof sof were-writing /wrote this to this until they finally nifgeSu (B2). met 3p.m.pi. (rec) 'For 10 years he and his brother in Moscow wrote to each other until they finally met 1 .

The tendency to use the pronominal form with the reciprocal is much smaller than it was with the reflexive. Research on children's use (Berman 1983, Berman & Sagi 1981) show that the HITPAEL is productive as a reciprocal form.

90

There is a difference between the reciprocal and the reflexive HITPAEL·, namely, that the reflexive can be derived both from PAAL and PIEL transitive verbs, whereas the reciprocal HITPAEL can be derived only from PAAL· transitive verbs. I suggest the following rule for deriving the reciprocal HITPAEL:17 (89)

RECIPROCAL HITPAEL FORMATION

input output

: :

meaning :

C.C.C. V Bl (xl)Ag (x2)Go C.C.C. V B7 (xl)Ag

"the action designated by the input predicate applies reciprocally between the individuals forming the group (xl) in the output predicate".

Active-intransitive and inchoative The main and most productive function of the HITPAEL· (according to my corpus) is to express the intransitive form, and to a lesser extent, the inchoative. In the literature this meaning of the HITPAEL· is often called 'Middle', following Latin and Greek grammars (for example Gesenius, 1910, Berman 1978). But analysis of the constructions in question shows that the verb in HITPAEL is in fact intransitive. For that reason I shall call it 'intransitive'. Consider the following illustration of the HITPAEL: (90)

(91)

...ceecaeiyhem gdolim, SeSuv eynam mitgorerim (B7) offsprings-their big, that-now not live 3p.pl. 'im hahorim. with the-parents 'Their children are big, they don't live any more with the parents.' pitom mistaber (B7) Seoto eru'a 'acmo miStakef (B7) suddenly turn-out that-same event itself reflects beSney 'itonim kmo sney eru'im Sonim in-two newspapers as two events different sehitrahaSu (B7) bli kol keSer beSney mekomot that-happened without any connection in-two places aberim. different 'Suddenly it turns out that the same event is reflected in two different newspapers as if there were two unconnected events in two different places'.

There are also cases where the HITPAEL· includes an extra feature, such as +momentaneous or +iterative, as in (92)

yaSav 'sat' 1 - hityaSev 'sat 1 +mom halax 'walked - hithalex 'walked' +iter.

These cases, however, are rare and irregular, and are included in the lexicon as idiosyncracies. In any case, here too the HITPAEL· is intransitive. It is, however, derived from the PAAL· not by valency reduction but by qualitative valency change. Cases such as this are discussed in chapter 5 section 5.5.1.4 . The complete Dictionary of Hebrew (Even-Shoshan, 1967) lists 269 verbs in the HITPAEL of which 165, as isolated verbs, have active- intransitive meanings. The text frequency of the intransitive HITPAEL is also considerably larger than that of the reflexive or reciprocal. In a

91

corpus of 89 predications with HITPAEL 80 were intransitive, 4 were reflexive, 4 were inchoative and 1 was reciprocal. Sentence (93) below illustrates the inchoative function of the HITPAEL·: (93)

macav briuto histaper (B7) meod laabrona. situation health-his improved greatly lately 'His health has improved lately a great deal.'

At first sight it is not always clear that the HITPAEL is actually derived by second argument reduction. For example, aitgorer in (90) is synonymous with gar, both meaning 'live', but there are also cases where the relation active-transitive —> inchoative (or resultative) is quite clear, as in (94) and (95) below: (94)a b (95)a b

sara cilma (Bl) otanu leyad hapesel. Sara photographed 3p.s.f. us near the-statue 'Sara photographed us near the statue', hictalamnu (B7) leyad hapesel. got-photographed lp.pl near the-statue 'We got photographed near the statue*. hapicuc nipec (B3) et kol hahalonot basviva. the-explosion blew-3p.s.m. GM all the-windows in-the-area 'The explosion blew to pieces all the windows in the area, hapicuc haya kol kax hazak sekol hahalonot the-explosion was so strong that all-the-windows basviva hitnapcu (B7). in-the-area were-blown-to -pieces 'The explosion was so strong that all the windows in the area were blown to pieces.'

Chapter 5, gives many cases where there is a semantic difference, i.e. a change in qualitative valency. It is the large group of roots occurring in the PIEL-PUAL-HITPAEL patterns, where the HITPAEL· is cearly derived from the PIEL, that bears out how productive second argument reduction is. The derivation of the HITPAEL· (-tr) can fall into two categories: a. change of qualitative valency (see also chapter 5 section 5.5.1.4). b. change of quantitative valency The changes of qualitative valency are illustrated here only briefly: (They are treated in chapter 5 section 5.5.4.1 .) 1. PAAL (-tr) -> HITPAEL· (intransitive, +mom or +iter) 2. PIEL -> HITPAEL (inchoative) as in: (96)

ciyer 'drew'

-

silhev 'made enthusiastic' -

hictayer

'came into view, 'appeared' histalhev 'became enthusiastic 1

This detransitivized HITPAEL will be derived by the following rule: (97)

DETRANSITIVIZED

input : output :

HITPAEL· FORKATION

C.C.C. V-B1, B3, B5 (xl)Ag (x2)Go C.C.C. V-B7 (xl)Proc

This is a productive rule. The HITPAEL forms resulting from qualitative change of valency are regularities, which should be listed in the lexicon.18 Note that there are now three formally similar rules for deriving the HITPAEL pattern: rule (86) for Reflexive, rule (89) for Reciprocal and

92

rule (97) for intransitive. All three reduce the second argument of verbs in the P AAL, FIEL, or the HIFIL. However, only the rule deriving intransitive can have all three binyanim as input; the reflexive has only PAAL and FIEL, and the reciprocal only PAAL as input. Furthermore, if for example a root is reflexive in the HITPAEL, it cannot form the reciprocal or the intransitive with this pattern, but will do so either by means of pronouns or by another form of paraphrase. (The rule deriving HITPAEL (-tr) has HIFIL as input only if the verb in question does not occur in the PAAL·.) The rules as given here have the following drawbacks: 1. The only generalizations that can be made are that if an input is HIFIL the output can only be intransitive (except if the HIFIL is a trivalent causative) and if the input is PIEL the output cannot be reciprocal. (But it can be reflexive or intransitive). 2. They do not exclude the roots which form the reflexive or reciprocal pronominally and not morphologically; 3. They do not prevent a root from being input to one of the rules which can actually only be input to another; i.e. they do not exclude the input of a root to reflexive formation, even if that root is reciprocal in the HITPAEL·. It is perhaps possible to formulate the restrictions on the input in terms of CSS's, but then the question arises whether the reflexive and reciprocal HITPAEL· can still be called rules, or whether they are only regularities, applying to a closed group of roots. This, however, is a methodological question concerning the definition of productivity and therefore falls outside the scope of the present discussion. (I will deal with this problem in chapter 6.) The HITPAEL· only meets the first of the two conditions, psychological reality and applicability to an open group. There is psychological reality, but it applies to a closed group. Thus in fact if the definition is to be strictly adhered to, these are not productive processes. This problem is discussed in chapter 6. 4.6.4 Doubl· reduction of valency An interesting question regarding valency reduction is whether there can also be a double reduction of valency, involving the reduction of three- place verbs to one-place. There are two types of trivalent verbs: a. basic predicates, like 'give'; b. derived predicates whose valency was increased, namely causatives. Three-place basic predicates are quite rare in Hebrew. The only ones known to me are saa 'put 1 natan 'gave', hiskir 'rented', horiS 'bequested* and biobil 'bequested'. Of these only bisJdr 'rented* can be used with one argument only: (98) (99) (100) (101) (102)

hadira huskera (B6) kvar. (from s.x.r. 'rent') the-flat was-rented already 'The flat was already rented'. *hasefer nitan (B2) (from n.t.n. 'give') the-book was-given *habayit hurafi (B6) (from y.r.s. 'inherit' *buraS does the-house was-inherited not exist) *harexus hunhal (B6) (from n.b.l. 'bequest, settle') the-fortune was-bequested *ha'iton husam (B6) (from s.u.m. 'put') the-newspaper was-put

93

Three-place verbs resulting from valency increase, i.e. causative verbs all follow the HIFIL pattern. Only 15 such verbs, (as given in Cole (1976b)) are known to me. A complete list is given below (103); in (103a) those which take two accusatives, or in FG terms Goal Markers et, and in (103b) those that take et and a dative Je): (103)a

b

heexil 'caused to eat, f e d 1 , hifSit 'caused to undress', birkid 'caused to dance', biSbi'a 'caused to swear 1 , biska 'caused to drink, watered', bilbiS 'caused to wear, dressed', he'evir 'caused to cross', hodi'a 'caused to know, announced', bixtiv 'caused to write, dictated 1 , bi&ai'a 'caused to hear', bikoa (yed'a) 'caused to acquire (knowledge)', bizkir 'reminded', bilbiS 'caused to dress', and bio'ία 'made pleasant'.

All of these verbs can undergo double valency reduction and take one argument only. Some of them even occur in 2 different binyania: the NIFAL and the HUFAL. There are various reasons for claiming that of these two only the HUFAL· is derived directly from the HIFIL by double valency reduction. (Whereas the NIFAL is derived from the PAAL; hiSbi 'a is an exception, as it does not occur in the PAAL.) These cases should not be confused with the PUAL/HUFAL vs. NIFAL pseudo- passives dealt with in section 4.6.2 . There both the NIFAL and the PUAL/HUFAL can be derived only from the PIEL or the HIFIL, whereas in the cases dealt with now, there is a PAAL basic pattern from which the NIFAL can be derived as argued in Junger (1985b). Here only one example is quoted: (104)a b

haaruha neexla (B2) the-meal was-eaten 3p.s.f. 'The meal was eaten.' haaruha huaxla (B6) the-meal was-fed 3p.s.f. 'The meal was fed (to)..'

Sentence (104a) relates to19 (105)

hi axla (Bl) et haaruha. she ate 3p.s.f. GM the-meal 'She ate the meal'.

whereas sentence (104b) relates to (106)

haah heexil (B5) et hahole et haaruha. the-male-nurse made-eat GM the-patient GM the-meal 'The male-nurse fed the meal to the patient'.

This process is in principle the same as in (79c), therefore it could in principle be described by a similar rule: 18 (107)

DOUBLE REDUCTION OF VALENCY

input : output :

C.C.C. V-B5 (xO)causer (xl)causee (x2)Go C.C.C. V-B6 (xl)Proc

94

Note that rule (107) applies only to those roots where the HIFIL expresses causativity, and is trivalent, i.e. only to the verbs listed in (103). But as it concerns only a rather small group of verbs, rule (107) is in fact not a predicate formation rule in the proper sense (as it is not productive), but only a regularity. (See also chapter 5 section 5.5.1.2). There are many other verbs where the HIFIL expresses causativity, but which have only two arguments: a causer and a causee. This is possibly because in most of these verbs the PAAL is intransitive, having valency 1, and the HIFIL· derived from it can have only valency 2, whereas the verbs in (103) are bivalent in the PAAL· pattern (except for hiSbi 'a 'cause to take an oath 1 , which does not occur in the PAAL). For example (108)a

b

hapo'alim hiäbitu (B5) et hanamal leSavu'a. the-workers made-strike 3p.pl GM the-harbour for-a-week 'The workers caused the work in the harbour to stop for a week.' hanamal huSbat (B6) lemeSex Savu'a. the-harbour was-put-on-strike 3p.s.m. for a-week 'The harbour was put on strike for a week'.

Such verbs can also undergo detransitivization, but that too follows rule (79c). Rule (79c) accounts for those verbs whose basic active forms is in the HIFIL, such as -v (109) hecil 'saved 1 —> hucal 'was saved' Second argument reduction, i.e. the HITPAEL pattern in the group of verbs with a trivalent HIFIL, can express reflexivity or reciprocity, but not intransitivity. (There are also verbs in this group which do not have a HITPAEL· at all.) The following table is a summary of the binyanim distribution of these verbs: (110)

HIFIMB5)

HUFAL(B6)

heexil 'made eat' hifsit 'made undress' hirkid 'made dance' hisbi'a 'made swear 1 hilbiS 'made dress' he'evir 'made cross' hizkir 'reminded' hikna 'made acquire 1 hiSmi'a 'made hear' hodi'a 'made know, announced'

PAAL(Bl)

NIFAL(B2) HITPAEL(B7)

+tr +

-tr +

+

+

+

-

+

+

+

-

?

+

+

-

-

(

+

Ref.

)

+

+

-

+

Ref.

+

+

-

-

+

+

-

+

+

+

-

+

+

+

-

+

Rec.

-

+

+

+

Rec.

95

Cont. table (110): HIFIMB5)

hera 'made see' hixtiv 'made write, dictate'

HUFAMB6)

PAAL(Bl) +tr -tr

NIFAL(B2) HITPAEL(B7)

-

+

+

+

Rec.

+

+

+

+

Rec

Like the NIFAL, the HITPAEL of the verbs in table (110) is not derived from the HIFIL, but from the PAAL. The PAAL is accepted as basic, and hence the HIFIL is derived from it. The only reason to derive the HITPAEL from the HIFIL· and not from the PAAL would be if it were semantically related to the causative and not to the active form. This, however, is not the case. 4.7 Summary This chapter dealt with the hypothesis that the relations between the binyanim in the Modern Hebrew verbal system can be described in terms of changes of quantitative and qualitative valency. The productive processes which were described in this chapter concern mostly quantitative valency changes, i.e. changes in the number of arguments. (The qualitative valency changes will be discussed in chapter 5.) . The productive processes in the binyanim system were examined, and the following description was suggested: 1. Valency increase: derivation of causative HIFIL from PAAL; increase of valency 1 or 2 of the PAAL predicate frame to valency 3 of the HIFIL predicate frame. 2. Valency decrease: a. first argument reduction: derivation of pseudo-passive NIFAL, PUAL and HUFAL from the transitive PAAL, PIEL and HIFIL respectively. b. second argument reduction: derivation of HITPAEL from PAAL·, FIEL and HIFIL. Note that actually in the strict sense these rules are not productive because they have no predictive capacity. It is not always the case that a PAAL·, PIEL or HIFIL pattern undergo first or second valency reduction, therefore on the basis of the occurrence of a certain root in either of these it cannot be predicted that the root will occur also in the reduced counterpart of the pair. It is possible to predict only that if a root occurs in the PUAL it occurs also in the PIEL·, if it occurs in the HUFAL it occurs also in the HIFIL pattern, and if it occurs in the HITPAEL it occurs also in the PAAL, PIEL· or HIFIL· (if the HITPAEL is not its only verbal pattern). One might claim that this should lead to deriving the PIEL·, HIFIL and PAAL· by valency increase, rather than deriving the PUAL·, HUFAL and HITPAEL by valency reduction. This option has been rejected because it covers only part of the cases (excluding all the roots which do occur in the PAAL, PIEL and HIFIL but do not undergo reduction). This would lead to an inconsistency in the derivational picture which cannot be sustained by any further independent argument. Two more general issues in the MH verbal system were also touched on: transitivity and passives. The traditional analysis of transitivity in Hebrew is based on Latin, where the verbs take accusative (or dative). When this analysis is transferred to FG transitivity in Hebrew is defined as a predicate frame of the type: pred V (xl)Ag/Fo (x2)Go

96

This definition, however, is too narrow. It excludes those verbs with valency 2 where (x2) is not a Go (but Compl). For that reason it was suggested to include (x2)Compl in the definition. But more importantly, an alternative approach was put forward to cover the whole question. Namely, by substituting the term valency for the necessarily fuzzy term 'transitivity'. Then, the valency of the verb would be advanced as the central feature of the description, allowing a verb with valency 2 to undergo valency reduction. This emphasis on valency, however, does not mean doing away altogether with transitivity. The question whether a verb is transitive or intransitive and the semantic function of its arguments (in particular of (x2)) remain important to a certain degree. In formulating the productive valency change rules, transitivity is important mainly with regard to causative constructions. In the next chapter we will see that transitivity also plays a role in analysing the Configurational Sub-Systems. Another consequence of the approach suggested here for the derivation of the NIFAL, PUAL and HUFAL is that HH neither has a passive in the traditional sense, nor in the FG sense of assigning Subject function to the Go. The constructions with these binyanim are detransitivized and not passive. They are, in other words, not the equivalent of their 'active' counterparts (with PAAL, PIEL and HIFIL) as they do not merely designate the same State of Affairs from another perspective, but rather a wholly different State of Affairs. A similar view is presented in Vet (1985) for French and Van Schaaik (1985) for Turkish. This subject has been dealt with in greater detail in Junger (1986), and will also be discussed in chapter 6.

Chapter 5

The Configurational Sub-Systems 5.0 Introduction Chapter 3 discussed the lexicon of Modern Hebrew and the representation of the roots and bioyanim. Chapter 4 dealt with the productive processes in the binyanim system using predicate formation rules, most of which could be stated in terms of quantitative valency changes of the predicate schemes associated with the binyanim (and some others in terms of changes in qualitative valency). In this chapter I will deal with the non-productive processes in the binyania system. These, as we shall see, are treated in the context of 'Configurational Sub-Systems' (CSS); this notion refers to a group of roots all of which occur in the same binyanim. (A detailed discussion of the CSS's follows in section 5.2) In addition, this chapter will test the two hypotheses underlying this research, which are briefly: 1. Host of the rules and regularities in the binyanim system can be described in terms of valency changes: quantitative and/or qualitative. 2. There is a considerably higher degree of regularity in the semanticsyntactic function of the binyanim and the relation between them within the CSS's than on the general level of the overall verbal system. Chapter 4 concentrated chiefly on hypothesis 1; this chapter concentrates on hypothesis 2. This hypothesis can be refined by dividing it into three connected subhypotheses, which will be checked separately. These are: (i) each Mayan has one syntactic-semantic function per CSS; (ii) each binyan has a consistent (quantitative) valency per CSS; (iii) within the CSS's the same valency changes concerning binyanim derivation occur as in the predicate formation rules. Ad (i> This is a stricter formulation of the general hypothesis regarding the consistency of the binyanim function in the CSS. Ad (ii)

This sub-hypothesis is based on the claim made in chapter 3 (section 3.1.1) that valency is at least as central a notion as transitivity in the Hebrew verbal system.

Ad (iii) If this sub-hypothesis turns out to be valid it reinforces the theory that the binyaoim are the organising principle of the Hebrew verbal system, while simultaneously giving the description of the verbal system presented in this study greater unity. It is important to note here that the syntactic-semantic functions of the binyanim are established on the basis of their use in concrete examples. Whenever possible these examples have been taken from a corpus,1 rather than being specially devised. (Examples devised by me are preferably used only in order to illustrate a point already established elsewhere.) The changes of valency in the derivational relations between the binyanim can in other words be based on actual usage. Otherwise, the assumption that

98

the processes in the binyanim system can be described in terms of valency would precede the analysis in terms of valency making the argument circular. (This circularity is avoided as actual usage is the basis for establishing valency.) Before proceeding to the CSS's themselves, the following methodological point should be noted: The consistency that will be reached holds by definition for relatively small groups of verbs, and is not comparable with the generality of the rules described in the previous chapter. 5.1 Theoretical background The survey of the approaches to the binyanim system in chapter 1 showed that there are two main streams on this issue: on the one hand there are those who see the binyanim as a syntactic system with certain regularities, featuring regular and predictable values as well as syntactically regular relationships; for example, Gesenius (1910), Blau (1972) ). Others, on the other hand, no longer wish to consider binyanim as the central syntactic system underlying the verbal system of Modern Hebrew. This view is taken chiefly by Oman (1969, 1971a, 1971b, 1971c) and Schwarzwald (1977, 1981a, 1981b, 1981c). An intermediate position is taken by Herman (1975a, 1975b). Berman wants to retain the notion of binyanim as the syntactic principle of the verbal system. She lists in detail not only the primary productive, but also the secondary functions of the binyanim (see chapter 1 section 1.3.1 table (12)) and suggests redundancy rules which would account for those productive relations forming the basis for morphological derivations. The main drawback of this approach is that as its rules only cover part of the possible relations, it cannot counter the objection of the 'lexicalist' approach that the rules do not account for the complete system. These approaches have in common that they see all the verbs as forming one system and try to account for the functions and features of the binyanim with regard to all the verbs. I believe that an attempt to account for the verbal system of Modern Hebrew should be based on the notion of the binyanim as the organizing principle. The main reason for this is that the binyanim are the chief (and sometimes the only) means of expressing syntactic features and syntactic relations like Active:Passive, Causativity, Reflexivity and Reciprocity. Furthermore, several of the binyanim's syntactic tasks in the system are still productive (e.g. the HIFIL for expressing causativity). Subdivision of the complete verbal system into sub-systems, in order to see whether the values and relationships of the binyanim are more regular within these sub-systems may provide a solution to this problem. A central question then is which principle the sub-systems should be based on? A first working hypothesis was to divide the whole inventory of roots occurring in verbal forms, into semantic groups: verbs of perception and cognition, verbs of movement in space, verbs of movement in time, verbs of emotion, verbs of action, verbs denoting states, etc. Some justification has indeed been found for regarding the verbs of perception and cognition as a separate group. Junger (1983) demonstrates that predications with verbs of perception and cognition belong to a specific group of State of Affairs, namely Position and State, but not to Action and Process. In addition, verbs of perception and cognition have a

99

specific predicate frame with the semantic functions Experiencer and Phenomenon. And finally, the so called 'passive* patterns of these verbs, the NIFAL and HUFAL should be seen in terms of detransitivization rather than as actual passives (i.e. resulting from Subj assignment to the Go). This in fact applies not only to verbs of perception and cognition but to the patterns NIFAL, PUAL and HUFAL in general. However, this approach could not be extended to other semantically based groups of verbs. Verbs of action, or of motion, emotion etc. do not exhibit any syntactic behaviour that would justify seeing them as a separate sub-system, nor are the binyania more regular within these groups. This working hypothesis was therefore rejected and replaced by a principle of classification into Configurational Sub-Systems, mentioned in section 5.0 . 5.2 The configurational sub-systems A configurational sub system (CSS) is a group of roots which occur in the same patterns. For example the roots r.t.h. 'boil 1 , n.u.s. 'escape* and S.y.t. 'sail' all occur in the PAAL, HIFIL and HUFAL· patterns; they thus belong to the same CSS; Similarily, the roots h.r.d. 'fear' and t.r.m. 'donate 1 which occur in the PAAL, NIFAL·, HIFIL and HUFAL· patterns, form another CSS, c.r.d. 'get hoarse', b.'.l. Own 1 , l.a.h. 'tire* occur in the PAAL, HIFIL and HITPAEL, and form yet another CSS. As stated above, the hypothesis tested in this chapter is that the binyania show a considerably greater consistency in their semantic-syntactic function and derivational relations within the CSS's than on the general level of the overall verbal system. Concretely, I want to establish whether the value of a certain binyan is consistent within a CSS. For example, in the CSS consisting of a.x.l.s. 'populate', m.m.5. 'realize', d.i.d.i. 'make scarce* the HITPAEL expresses in all cases indication, whereas in the CSS consisting of c.y.d. 'supply* and S.h.r.r. 'free* the HITPAEL expresses reflexivity. This chapter aims at finding out whether the existence of such regularities holds good for all the CSS's. Thus, for example, the NIFAL pattern which if regarded generally can express intransitive, pseudo-passive, inchoative (and in a small number of cases reflexive and reciprocal), would in a certain configurational sub-system be consistently pseudo-passive, and in other sub-systems consistently intransitive. Finding such consistencies in the CSS's is important, because they could serve as the basis for organizing the lexicon. The prototypical cases of binyanim functions and relations, which have been discussed in chapter 4, can be dealt with in terms of predicate formation rules. These rules, however, hold only for about 30% of all the roots (see the big CSS in table (16) further in this chapter). The ability to organize all the remaining roots into CSS's exhibiting some regularity enables the formulation of rules for these roots too. Within the framework of F6, rules which are not productive are part of the lexicon. For this reason, CSS's can operate as the organizing principle for the lexicon of MH verbs. In addition, it will be interesting to see whether the CSS's can provide any explanations for why certain roots do not occur in certain binyania. Providing a clear picture of the distribution of the roots in the CSS's and of the productive and non-productive functions of the binyania in these CSS's could provide a first step towards explaining gaps in the system. The distribution of roots in the various binyania shows only occurrence and cooccurrence, not any causal relationship. But if this

100

distribution is based on any semantic, syntactic (or other) principle(s), than this principle can also serve as the (at least partial) explanation of the occurrence and therefore also the non-occurrence of the roots in question in certain binyanim. It it turns out that in certain CSS's the NIFAL pattern for example, expresses always pseudo-passive which is derived from the same root in the PAAL pattern, then the absence of some roots in the PAAL· pattern, might also explain their absence in the NIFAL· pattern. Note that this way of explaning gaps is always limited to specific CSS's and roots, and does not apply to the whole system. (If we take the example just mentioned, there may also be CSS's featuring the roots both in the PAAL (+tr) and in the NIFAL, whereby the NIFAL· expresses another function, say inchoativeness.) It is important to add that we cannot expect the relation between syntactic value and CSS to be 1:1 . Simply, because the number of configurational sub-systems is much larger than the number of syntactic features and relations expressed by the binyanim. Furthermore, there are binyanim which throughout different CSS's consistently express only one syntactic feature, like the PUAL and HUFAL which express only pseudopassive. 5.2.1 The procedure for establishing the CSS'n Before listing the actual configurational sub-systems that were found, several notes must be made regarding the methods by which they were established. The primary source for the roots was Barkali's (1980) Complete Table of Verbs (a revised version of Barkali 1965). This is an alphabetical list including all the roots in Hebrew occurring in verbal patterns, and the binyanim in which they are featured. The list is complete in the sense that it is not limited as regards historical time, register or frequency. It includes all the roots, from Biblical Hebrew (1200- 130 B.C.) through Hishnaic and Talmudic Hebrew (400 - 500 A . D . ) , middle and late medieval, right up to contemporary Israeli Hebrew. It includes all the verbal patterns in which a certain root has been encountered, in written or oral form, whether rare (occurring only once or twice) or frequent and clearly productive. This whole table has been classified by means of a computer into groups, according to occurrence in specific binyanim. Mathematically there could be 127 possible combinations between the seven binyanim. In the classification based on Barkali (1980), however, there are only 102, including 7 groups of roots occurring in one binyan only. Thus there are 95 configurational sub-systems with 2 or more binyanim. In the complete list of all the roots from Biblical Hebrew up to Modern Hebrew (Barkali, 1980), there are in total 3181 roots. The CSS's were established by the following procedure: (i) the complete list in Barkali (1980) is the 'population 1 ; (ii) all roots occurring in one to seven binyanim are 'sub populations'; we get the following 7 sub-populations: a. all the roots occurring in one binyan; (414) b. all the roots occurring in two binyanim; (708) c. all the roots occurring in three binyanim; (664) d. all the roots occurring in four binyanim; (308) e. all the roots occurring in five binyanim; (334) f. all the roots occurring in six binyanim; (339) g. all the roots occurring in seven binyanim; (414)

101

(iii)

each sub-population in (b)- (f) can be further broken down according to the combinations of binyanim in which its members occur. For example, in sub-population (b) we can find the following combinations: PAAL-PIEL, PAAL-HIFIL, PIEL-HITPAEL, etc. Or, in sub-population (d) we can find for example the combinations PAAL NIFAL HITPAEL HIFIL, or PIEL PUAL HITPAEL, HIFIL, or NIFAL PIEL PUAL HITPAEL, etc. Every combination of this type is a configurational sub-system (CSS). Not all the CSS's are studied in detail. A limit of 1% of the relevant sub population was established, and whichever group fell under it was left out of consideration. Thus, in the sub-population of two bioyaoim (group (b) above), the limit of 1% is 7; consequently the CSS's with less then seven roots were excluded. In the sub-population of four binyanim the limit of 1% is 3, so all the CSS's with less than 3 members were excluded. This limit was chosen, because a group smaller than 1% is considered too small to represent any significant regularity. A varying limit of 1% of the sub-population was preferred to a constant limit of say 1% or 0.5% of the whole population for the following reason: a limit of 16 roots (0.5%) represents a different value within each sub-population, being a different percentage of the sub-population. In a sub-population with 708 members it has a different value than in a sub-population with 334 members. A constant limit would have been suitable only if all the sub-populations would have been more or less of the same size. 5.2.2 Methodological remarks During detailed study of the roots and the CSS's it turned out that many roots and combinations of root + bioyan listed in Barkali (1980) were not known to me. Therefore, the New Hebrew Dictionary by EvenShoshan (1967) which also lists all the roots occurring in Modern Hebrew has been added, to allow counterchecking. Even-Shoshan, like Barkali, does not exclude obsolete forms, but beside each pattern of a root he mentions its first recorded date or period of occurrence. Of course, this is no indication whether the form in question is still part of MB or not. It only gives certainty regarding those rare forms, not known to me but listed as 'Modern Hebrew, literary', which might have been excluded otherwise. As for those forms from Barkali which were unknown to me, it turned out that many were not mentioned at all in Even-Shoshan's dictionary. Thus the following three sources were used: a. Barkali (1980) b. Even-Shoshan (1967) c. researcher's knowledge (on the level of a near native-speaker of Hebrew) and a random corpus compiled for this study consisting of 561 sentences taken from spoken Hebrew, newspapers and journals, and contemporary literature. (For details on the corpus see note 1 to chapter 4.) When a certain form was unknown in two of the three sources (Barkali, Even-Shoshan, or the author's knowledge), it was taken off the list of roots which formed the 'population'. Those roots which did not appear in Even-Shoshan (1967), or did appear but are now obsolete were removed. There was quite a number of these cases. For this reason, the whole population was re-established. The new population was classified into sub-populations and CSS's by means of the computer. Then the floating limit of 1% of the sub-populations was applied again in order to obtain the final list of CSS's. This chapter only gives the results of the final list of CSS's.

102

The reestablished population, which was the basis for the CSS's is considerably smaller than Barkali's. Barkali (1980) gives a list of all the roots in Hebrew, biblical, post-biblical and modern. This research, however, is synchronic. Many linguists studying Modern Hebrew claim that the synchronic view cannot be completely separated from the diachronic one, because many phenomena in Modern Hebrew can be explained only in terms of the diachronic development of the language. Regarding the verbal system, for example, they claim that the frequent occurrence of apparently arbitrary root distribution in the binyaaim system is explained by the simple fact that when a change took place the binyan most suitable for expressing it was already Occupied*. (This explains for example part of the cases why the HIFIL, whose primary function is causative, is active-transitive.) In other words, a synchronic phenomenon has a diachronic explanation. Nevertheless, I think that it is still possible to limit the present research to a synchronic study of Modern Hebrew. Even Shoshan's list of binyanim, like Barkali's, is not limited by historical time, register or frequency. Unlike Barkali, Even-Shoshan indicates the historical period in which a certain root started to be used in a certain pattern. So for example, the root s.u.t. 'move' occurs in the PAAL, NIFAL, HIFIL· and HUFAL patterns; the PAAL, NIFAL· and HIFIL are mishnaic, and the HUFAL is modern literary Hebrew. Or the root k.S.v. 'listen' occurs in the PAAL, NIFAL FIEL HIFIL and HUFAL; its use in the PAAL· and HIFIL· is biblical, in the NIFAL, PIEL and HUFAL it is medieval. However, it is difficult to draw any further conclusions from this information. (All it means is that the oldest source in which the compiler of the list encountered the form in question is biblical, mishnaic, medieval, etc.) It does not indicate whether the form in question is used in present day Hebrew. (With the exception of the rather limited number of verbs listed as originating in modern literary Hebrew.) It is impossible to establish with absolute certainty which roots are used in which binyanim in MH on the basis of the present study, because there is no representative corpus of MH. By representative corpus I mean something like the corpus compiled in P.C. uit den Boogart Word Frequency in Spoken Dutch (1975). This corpus includes 720,000 words from various written sources, as well as recordings of conversations between people of different socio-economic classes. No such corpus exists for present day Hebrew, and compiling and analysing one was clearly beyond the possibilities and scope of this study. It is also possible that some of the new colloquial uses which are permanently entering MH, are not represented in this study. This problem, however, is of a more general nature, and inherent in the attempt to pin down a language, which is by nature a dynamic phenomenon. The possibility of using an arbitrary sample was rejected, because owing to the nature of this population an arbitrary sample might have excluded productive phenomena, or given a distorted picture (present nonproductive ones as productive). Therefore, from a methodological point of view the best option was to check all the roots. As mentioned already, there is no corpus on which I could rely, and the comprehensive Hebrew dictionaries, like Even-Shoshan (1967), or Alcalay (1969) include old and obsolete forms, without marking them or letting the user know whether these forms are still in use in MH. In cases where I suspected a form to be obsolete, I had to rely on my knowledge of Hebrew. Whenever I encountered a word that I did not know, but that made sense, I left it in the population. The reason is that no speaker of a language knows every single word in the langauge. (According to Schwarzwald (1981b) an average adult has a vocabulary of approximately 25,000 words.) Despite these

103

limitations, I beleive that the error margin in this study is not beyond the tolerable, and that the data analysed are synchronically relevant. Analysis of certain texts, or spoken language, can at most indicate frequency of use, from which at best conclusions can be drawn with regard to the productivity of certain binyania in Modern Hebrew. But a complete list requires a large corpus featuring representative material from all registers, and such a corpus does not exist. Another possibility would have been to use frequency tables. The largest frequency table for MH is Balgur (1968). According to the introduction it contains 4224 lemmata, which recur in texts, adding up to 185,899 words, from a total of 200,000 words in all the texts. (However, it is not explained clearly what is meant by lemmata; the list consists of concrete words, i.e. roots + miSqalim and roots + binyania.) The texts used for this list are of the following types: school books, youth literature and youth journals. Thus, in fact, Balgur gives a picture of what can be called 'basic Hebrew 1 which means that its use as the basis for this research would have raised several problems. The great majority of the roots in this list, for example, are of the type that occur in all 7 binyanim (even if they are not listed in all of them on the frequency list). But it has been shown by Schwarzwald (1981) that only some 2.3% of the roots occur in all 7 biayaoim.2 Therefore, Balgur's list used as a corpus would give an unrepresentative picture. Moreover, this research aims at describing some aspects of the verbal system of MH, not of 'basic MH'. In other words, the frequency factor will always yield an innaccurate picture. A methodological problem of a totally different nature was the disagreement between Barkali (1980) and Even-Shoshan (1967) over the roots and the patterns in which they are supposed to occur. Consequently, the same root often had a different representation in the two sources. These differences fall into one of the following 3 types: A. different representation of the root B. the quadriliteral roots C. homophonous roots A. Different representation of the root The two sources disagreed sometimes on what was understood to be the root representing a certain semantic field. This difference is evident in the spelling, and may be systematic, as illustrated in (i), or arbitrary as illustrated in (ii). (i) Systematic differences The systematic differences in representing the root always regard the first or second consonant of the root, when these are /y/ or /v/. This is illustrated by (la) and (Ib) respectively: (l)a b

Barkali y.'.d. n.y.d. n.y.b.

Even-Shoshan v.'.d. 'gather 1 n.u.d. 'move' n.u.h. 'rest'

These differences in spelling, however, did not influence the number of roots. The same roots occurred in both lists, only with different spellings. I.e. in Barkali (1980) there is an entry y . ' . d . , its equivalent in Even-Shoshan (1967) is 'see v . ' . d . '. (ii) Arbitrary differences These differences are called arbitrary here, because they do not concern a systematic exchange of one root consonant for another. For example:

104

(2)

Barkali s.h.t. s.r.k. t.a.v. t.a.v. ffl.n.h.

Even-Shoshan s.h.t. 'squeeze' (1st root-consonant) 3 s.r.k. 'comb' (1st root-consonant) t.'.v. 'disgust' (2nd root consonant) a.v.h. 'wish* (all 3 root-consonants) n.u.h. 'rest* (1st and 2nd root consonants)

This difference in spelling has also lead to a difference in the number of roots on the list. Thus according to Barkali k.t.t. and k.v.t. 'quarrel* are two different roots, whereas according to Even-Shoshan they are two different ways of spelling the same root. Unlike EvenShoshan (1967), Barkali (1980) only lists the roots without adding any further information. Even-Shoshan gives in the entry k.t.t. the meaning 'quarrel', but since Barkali lists only k.v.t., there is no way of knowing whether he meant the same root (i.e. k.v.t.). The root of words like Jtrata 'a quarrel' or hitkotet 'quarelled* (reciprocal) is k.t.t.; Even-Shoshan in the entry for k.t.t. refers to k.v.t., but there is still a small chance that Barkali meant a different root, as can be seen from other examples, where according to Even-Shoshan we have two different spellings of the same root, but according to Barkali we have two different roots. This means that there is no way of ascertaining conclusively the number of roots concerned. The following table illustrates several cases of such differences in the registration of roots. In Even-Shoshan (ibid) the two spellings represent the same root, and the entries refer to each other. The entry for k.d.d. 'cut out a piece' says "see k.d.h.". In Barkali (1980) only one of these representations can be found. Höre often, however, both spellings are listed. But it is the distribution of the roots, not the actual meaning of the words, which indicates the difference. (3)

Even-Shoshan k.d.d. =k.d.h. k.u.n. =k.y.n. n.c.a. =y.c.a. n.c.g. =y.c.g. n.c.v. =y.c.v. y.r.'. = r . ' . ' . y.t.v. =t.u.v. r.z.n. =r.z.h. r.z.k. = r.s.k.

Barkali no k.d.d. k.d.h. Bl k.u.n.'get ready' B3, B7; k.y.n. '?' B3, B7 y.c.a. 'go out 1 Bl, B3, B4, B5, B6 n.c.a. '?' Bl y.c.g. 'represent* B2, B3, B4, B5, B6, B7 no n.c.g. n.c.v. = y.c.v. y.r.'. Bl r . ' . ' . 'befriend' Bl, B2, B3, B4,B5 B6, B7 y.t.v. Bl, B5, B6 t.u.v. Bl, B5, B6 r.z.h. 'thin* Bl, B2, B3, B6 r.z.n. '?' B7 r.z.k. B3, B4, B5, B7 r.s.k. Bl, B3, B4, B7

This table shows that there are fewer entries listing roots and their verbal derivatives in Even-Shoshan's dictionary than in Barkali's table. Even-Shoshan gives the following statistical information regarding the roots in the dictionary:

105

Source Number of roots Bible 1486 Talmud and Nishna (till 5th cent.) 664 medieval literature (till 17th cent.) 164 Modern Israeli Hebrew 329 Total 2643 That means, 538 (16.9%) roots less than the number listed by Barkali. The number of CSS is also smaller than what was found on the basis of Barkali (ibid); instead of 95 CSS's there were only 85 with two or more binyania. B. Quadriliteral roots Another large group of roots where there is disagreement between the two sources is formed by the quadriliteral roots. There are two types of quadriliteral roots: (i) of Biblical origin (ii) Modern Hebrew, usually denominal or deadjectival The former group consists largely of two consonants, which are reduplicated, such as (4)

g.l.g.l. S.f.S.f. c.l.c.l. p.r.p.r.

'wheel, turn around' 'rub 1 'ring' 'flutter 1

Note that these roots, and the words derived from them, are mostly onomatopoeic. These roots are still widely used in Modern Hebrew. The second group of MH innovations can be clearly distinguished from the first one, because all its consonants are different, and are semantically very clearly denominal or deadjectival. The disagreement between Barkali and Even-Shoshan with regard to these roots is whether to represent them separately, as in Barkali, or include them under the triliteral, old existing roots from which the noun or adjective in question is derived, as in Even-Shoshan. Here are several illustrations of the 'primary root', the noun derived from it, and the 'secondary root', extracted from this noun: (5)

Even-Shoshan (1st root) b.z.r. 'return 1 k.f.l. 'multiply' '.v.d. 'work' s.x.n. 'dwell 1

->

Noun -> Sibzur 'reconstruction' Sixpul 'stencil' 'eved 'slave' miSkaa 'dwelling 1

Barkali (2nd extracted root s.h.z.r. 'reconstruct' S.x.p.l. 'stencil' s.'.b.d. 'enslave' m.s.k.n. 'setle 1

Even-Shoshan also includes such quadriliteral roots, like a.f.S.r. 'possible* from efSarut 'possibility', but much less frequently than Barkali. The majority of Barkali's quadriliteral roots have the following distribution: PIEL- active transitive, PUAL· - 'passive 1 HITPAEL - middle. In Modern Hebrew there is a strong tendency to use the PIEL· pattern for new or borrowed verbs. (This has been noted among others in Bolozky (1978), cf. chapter 4, section 4.6).

106

C. Homophonous roots Hebrew features many homophonous roots. These are roots which are phonologically identical, but nevertheless form two independent roots, as they refer to two totally different semantic fields. In Barkali (1980) they are listed as one root. For example (6)a

Barkali: p.z.z. (Bl, B2, B3, B4, B6, B7) Even-Shoshan: p.z.z. 'light 1 (B6, B7) p.z.z. 'dance' (Bl, B2, B3, B4)

b

Barkali: c.r.f. (Bl, B2, B3, B4, B5, B6, B7) Even-Shoshan: c.r.f. 'add, join, model'1 (B3, B4, B5) c . r . f . 'add, join, model (Bl, B2, B6, B7)

c

Barkali: k.u.m. (Bl, B3, B4, B5, B6, B7) Even-Shoshan: k.u.m. 'keep* 1 (B3, B4, B5) k.u.m. 'raise (Bl, B6, B7)

In Even-Shoshan (1967) such roots have two separate entries. This means that in the original classification for CSS's (which is based on Barkali (1980) the root p . z . z . , for example, occurs in the sub-population of 6 binyanim PAAL, NIFAL, FIEL, HIFIL, HUFAL and HUFAL. In the final list such roots were removed from the CSS's they were originally assigned to and reclassified according to the binyanim listed in EvenShoshan 's dictionary. This occurred mostly in CSS's with 4-7 patterns. After checking the two sources against each other, and using the writer's personal knowledge of HH, the final number of roots in the various CSS's differed greatly from that established initially. Therefore, the resulting data were recategorized in the same way as for the list from Barkali (1980). The rest of this study will only take this new 'population' and the CSS's it contains into account.

5.2.3 General picture of the root-binyanim distribution This section will deal with some general facts that emerged from the classification of the roots into 'sub-populations' and CSS's. They can provide interesting indications as to the direction in which the binyanim system is developing. The new 'population* consists of 2452 roots, 4 divided into the following sub-populations: (i) roots in one binyan : 263 (ii) roots in two binyanim: 574 (iii) roots in three binyaninr. 691 (iv) roots in four binyanim: 320 (v) roots in five binyanim: 257 (vi) roots in six binyanim: 185 (vii) roots in seven binyanim: 165

107

This distribution can also be represented in the following graph: (7) frequency 700 650 600 550 500 450 400 350 300 250 200 150 100 50

number of biayaaia The biayaaia distribution in this graph is bell-shaped. The highest number of roots occur in 2-4 patterns, and fewer in 1, 5, 6 and 7 patterns. The CSS's of the roots occurring in the largest sub-population, occur in the following binyaaim: PIEL (+tr/c) - PUAL (pp) - HITPAEL (-tr). Owing to their high frequency with these functions these three biayanim are likely to be the most productive in MB. The system as a whole, consisting of all seven patterns has a low frequency and seems to be in decline. The productivity of the PIEL-PUAL-HITPAEL has also been pointed out by Berman (1981) for child language, and by Bolozky (1978). Interestingly, this classification reaches the same results. A further implication is that the biayaaia functions and the relations between them as represented by these CSS's should be productive too.

108

5.3 The Configurational sub-systems The lists of the sub-populations and the CSS's they contain in the final population were the following: (i) l biayan (total 263 roots) PAAL (85 = 32.3%)

NIFAL (8 = 3.04 %) PIEL (101 = 38.4 %) PUAL (0) HIFIL (31 = 11.8%) HUFAL (0)

HITPAEL (38 = 14.4%) These roots can be divided into two groups: 1. denominal 2. without category The first group consists of roots which originate from clearly nominal forms, either originally Hebrew or borrowed. For example (8)

noun ma'arav 'west 1

derived root -> m . ' . r . v .

verb -> hitma'arev 'became westernized' 'axSav 'now* -> '.x.S.v. -> 'ixSev 'actualized' cmarmoret 'shudder* -> c.m.r.m.r. -> hictamrer 'shuddered' flirt ' f l i r t ' -> f.l.r.t. -> flirtet 'flirted'

These roots occur mostly in PIEL; many of them are quadriliteral (65), and some even quintiliteral (5). The second group consists of roots which are not extracted from a word, and are therefore without category. Most of these are rare, and some only occur in the one verbal form. This sub-population consists of the following CSS's listed in table (9) below. The CSS's are numbered as follows: the first digit, 1, 2,3 etc., denotes the distribution of roots in biayaaim, regardless of the semantic-syntactic functions of the binyanim, i.e. CSS's; the second digits, .1, .2, .3 etc denote a further sub-division into functional units, within which the binyanim have only one semantic-syntactic function. That is, all the vertical lines with the same first digit belong to the same CSS, whereas the different second digit lines show the functions of the binyanim in the CSS's. The main numbers of the CSS's (i.e. the first digit) are continuous throughout all the sub-populations to permit easy reference to the CSS's in later sections. (9)

CSS 1 32.3 2 3.1 3.2 4 5.1 5.2 6 7

Bl -tr

B2

B3

B4

B5

B6

B7

-tr +tr -tr χ

+tr -tr χ

-tr

Examples n= c.l.'. 'limped' n.x.a. t.r.t.r. h.s.s. — r.i.h. m.'.d. ~ n.c.l. '.c.l.n.

nt= % 85

'sad 1 8 3.04 'bother' 100 'hesitate 1 1 101 38.4 0 'smell' 30 'stumble 1 1 31 14.4 Ο 'excuse' 'lazy' 38 11.8

legend: (n =number; nt= total number in CSS; % =percentage of sub-population)

109 Note that no verbs ever occur exclusively in the PUAL (B4) or the HUFAL (B6) patterns; these patterns, in other words, cannot be basic. This is supported by table (10) which shows of no CSS's consisting of roots that occur exclusively in these two patterns. (ii)

2 biayaaia (total 574 roots, 1%= 6) 1. 2. 3. 4. 5. 6. 7. 8. 10. 11. 12. 13. 14.

FIEL PUAL (22 = 38.5%) FIEL HITPAEL (97 = 16.9%) HIFIL HUFAL· (84 = 14.6%) PAAL· NIFAL· (79 = 12.9%) PAAL· HIFIL (32 = 5 . 6 % ) PAAL· PIEL (25 = 4.0%) PAAL· HITPAEL· (14 = 2.4%) PIEL· HIFIL· (9 = 1.5%) NIFAL· HIFIL· (3 = 0.5%) HIFIL· HITPAEL· (3 = 0.5%) PUAL HIFIL· (2 = 0.3%) PAAL· PUAL· (2 = 0.3%) NIFAL· PIEL· (1 = 0.2%)

Some possible combinations which are theoretically possible but never occur are: a. PUAL· - HUFAL· b. PUAL· - HITPAEL c. HUFAL· - HITPAEL· d. PUAL· - NIFAL e. PAAL· - HUFAL· Cases (a -d) are probably not used owing to the lack of a bivalent basic pattern. Both patterns are univalent, and hence cannot be derived from each other. (We saw in table (9) that the PUAL· and HUFAL· cannot occur as only pattern.) They would be most likely to occur together when there is a bivalent pattern in the CSS. The combination in (e) does not occur probably because although in principle it could be possible to derive a root in the HUFAL pattern from the same root in the PAAL· pattern (when the latter is transitive) by valency reduction, in practice the HUFAL is always derived from the HIFIL. (See also chapter 4 section 4.6.3 .) This sub-population consists of the following CSS's: (Those under the 1% limit are not listed)

(10) CSS Bl 8.1

B2

B3 +tr

B4 pp

8.2

c

pp

9.1 9.2

+tr c

10.1 10.2 11.1 -tr

-tr (ref) 12.2 +tr pp

B5

B6

c pp +tr pp

B7

Examples n= nt= % t.x.n.t. 'program' n.t.r.l. 'neutralize' 212 a.v.t.h. 'make safe' 9 221 38.5

-tr i

s.m.l 'symbolize' s.h.r.r. 'dizzy 1

80 17 97 16.9

g.r.l. p.g.z.

'draw lots' 'bomb'

71 13 84 14.6

a.n.k.

'sigh'

b.h.n.

'test 1

2 77 79 12.9

110

Cont. table (10) CSS 13.1 13.2 13.3 13.4

Bl -tr +tr -tr i

B2

14.1 +tr

B3

B4

B5 B6 +tr +tr c c/i

B7

+tr (int) +tr

14.2 -tr

-tr -tr +tr -tr

n= 6 3 22 1

S.b.r. 'break 1

23

h.t.r. 'undermine'

15.1 +tr 15.2 -tr 16.1 16.2

Examples g.v. 1 . 'die' m.r.h. 'disobey 1 n.d.m. 'quiet' c.h.v 'yellow'

c c

a.v.h. 'want' g.u.r. 'live 1 h.p.a. 'protect 1 d.l.g. 'jump 1

nt= %

32

5.6

2

25

4.0

1 13

14

2.4

2 7

9

1.5

The following points should be noted regarding table (10): 1. The terms +tr (transitive), -tr (intransitive), c (causative), pp (pseudo-passive), ref (reflexive), rec (reciprocal) and i (inchoative) are used here and throughout the whole study according to the definitions in chapter 4 sections 4.1.1, 4.5.1, 4.6.2 and 4,6.3 . 2. The gap between the largest CSS and the second largest is enormous, being more than 100%. This means that when roots occur in two binyanim the frequency of occurrence in FIEL and PUAL is significantly higher than in the other combinations. Also, in so far as it is possible to extrapolate from frequency to productivity, this CSS (i.e. these patterns) are productive in MH. 3. Within the CSS 1 s the binyanim functions are not consistent. As we shall see this applies to most CSS's with more than two binyanim too. This point will be discussed in detail in section 5.4 . 4. There is no CSS consisting of PUAL (B4) and HUFAL· (B6) only which derives these two patterns by first argument reduction, for they cannot occur without a bivalent pattern acting as the source. This fact supports the analysis given in chapter 4. It is probably for the same reason the CSS's consisting of NIFAL (B2) and PUAL (B4) or of NIFAL (B2) and HUFAL (B6) do not occur either. 5. Unlike the PUAL· and the HUFAL the HITPAEL ( B 7 ) , which is also analysed in chapter 3 as a pattern resulting from valency reduction, can occur in a CSS with another univalent pattern only; namely in CSS 15.2 with the PAAL (-tr). The derivational relation here is one of qualitative valency change and will be discussed in section 5.5.1.4 . (iii)

3 bioyanim (total 691 roots, 1%=7) 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

PIEL PUAL HITPAEL· (432 = 62.5%) PAAL HIFIL HUFAL· (50 = 7.2%) PAAL NIFAL· HIFIL (46 = 6.6%) PAAL NIFAL PIEL (24 = 3.5%) NIFAL· HIFIL· HUFAL· (21 = 3.0%) PAAL· PIEL· PUAL· (19 = 2.7%) PAAL PIEL· HITPAEL (19 = 2.7%) PAAL NIFAL HITPAEL (16 = 2.3%) PAAL· HIFIL· HITPAEL· (11 = 1.6%) HIFIL HUFAL HITPAEL· (9 = 1.3%) PIEL HIFIL· HUFAL· (7 = 1%)

Ill 12. FIEL PÜAL HIFIL (7 =1%) 13. 14. 15. 16. 17. 18. 19. 20.

PAAL NIFAL HUFAL (5 = 0.7%) NIFAL FIEL PUAL (5 = 0.7%) FIEL HIFIL· HITPAEL (5 = 0.7%) PAAL· NIFAL· PUAL· (2 = 0.3%) PAAL PUAL· HITPAEL (l = 0.1%) PAAL· PUAL· HIFIL· (l = 0.1%) NIFAL· FIEL HITPAEL· (l = 0.1%) NIFAL FIEL HIFIL· (l = 0.1%)

Combinations which could occur in principle, but do not actually occur

are:

a. NIFAL· PUAL· HUFAL b. NIFAL PUAL· HITPAEL c. PUAL HUFAL HITPAEL

Such CSS's would consist solely of univalent or reduced binyanim. This, as we saw in tables (9) and (10), is not possible; there must be a bivalent binyan which can undergo valency reduction. This sub-population consists of the following CSS's (those falling under the 1% limit are not included in the table):

(11) CSS

Bl

B2

B3

17.1 17.2 17.3 17.4 17.5

B4

B5

B6

+tr pp -tr pp c pp +tr pp -t-tr pp

18.1 -tr 18.2 +tr 18.3 -tr

c c c

B7 -tr -tr i ref rec

pp pp pp

Examples n= nt= % s.f.s.f. 'rub 1 342 h.l.h.l. dribble' 5 a.x.l.s. 'populate' 69 s.h.r.r. 'free 1 14 p.y.s. 'reconcile' 2 432 62.5 d.h.r. 'drive fast' 46 n . ' . m . 'pleasant 1 2 b.v.a. 'come' 3

51 7.2

k.d.h. 'dig' 2 p.z.l. 'to be cross-eyed* 2 g.v.z. 'disappear* 41

45

6.6

24

3.5

19 1

21

3.1

p.z.z. 'sing, dance' 4 15 r.h.t. 'furnish 1

19

2.7

h.l.x. 'walk' '.s.k. 'deal'

18 1

19

2.7

m.s.r. 'give' r.t.m. 'put in harness' r.k.k. 'spit'

12 17

2.3

19.1 +tr 19.2 -tr

pp -tr

c +tr

19.3 -tr

-tr

c

20.1 +tr

pp

+tr (inten)

c.r.m. 'disturb'

20.2 -tr

-tr

+tr (iter)

x.l.a. 'finish

21.1 21.2

-tr i

c c

22.1 -tr 22.2 +tr

+tr pp +tr pp

23.1 -tr 23.2 +tr

+tr +tr

m.r.c. 'zeal 1 h.l.d. 'rust'

PP PP

24.1 +tr pp 24.2 +tr i

-tr -tr (dur) -tr i

24.3 -tr

-tr

-tr/pp

21

3 2

112

Cont. table (11): CSS Bl B2 25.1 -tr

25.2

B3

+tr +tr +tr (iter)

26.2 +tr 26.3 -tr

B5 c

B6

c

+tr

26.1 -tr

B4

27.1 27.2

B7 Examples -tr g.s.s. 'be coarse* (sudd) -tr h.g.r. 'limp'

c +tr +tr (iter)

-»-tr(i) pp -tr c pp -tr +tr

28

PP

n= nt= % 10 1 11 1.6

s.h.k. 'discolour' g.h.k. 'burp 1

6 2

r.c.d. 'glitter'

1

r.s.m. 'impress' l.a.m. 'nationalize'

6 1

9

1.3

d.b.r. 'destroy insects'

legend: The abbreviations denoting aktionsart are used as follows: sudd = sudden, iter = iterative, inten = intensive The following points are noteworthy about table (11): 1. As in the previous two tables there are no CSS's consisting of reduced binyanim only (B2, B4, B6, B7). There are, however, CSS's consisting of univalent binyaoim only; these are numbers 17.2 and 24.3. These CSS's contain candidates for the basic pattern, namely FIEL and PAAL respectively. The other binyanim are derived from them by qualitative valency change. (See also section 5.5.1.4.) 2. There is only one CSS in which none of the bioyania has more than one function: number 12. (iv)

4 biayania (total 320 roots, 1%=3) 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

PAAL NIFAL HIFIL HUFAL (79 = 24.5%) PAAL· NIFAL PIEL PUAL (50 = 15.5%) PAAL· PIEL PUAL HITPAEL (30 = 9.3%) PAAL NIFAL HIFIL HITPAEL (23 = 7.1%) PAAL NIFAL FIEL HIFIL (20 = 6.2%) PIEL· PUAL· HIFIL HITPAEL· (19 = 5.9%) PAAL· HIFIL· HUFAL HITPAEL· (17 = 5.3%) PAAL· NIFAL· PIEL· HITPAEL (15 = 4.6%) PAAL PIEL HIFIL HITPAEL (13 = 4.0%) NIFAL PIEL PUAL· HITPAEL (13 = 4.0%) PAAL PIEL PUAL· HIFIL (10 = 3.1%) FIEL PUAL· HIFIL· HUFAL (8 = 2.5%) NIFAL· PIEL HIFIL HUFAL (6 = 1.9%) NIFAL PIEL PUAL HIFIL (4 = 1.2%)

15. 16. 17. 18. 19. 20.

FIEL HIFIL HUFAL HITPAEL· (3 = 0.9%) PAAL NIFAL PUAL· HIFIL (2 = 0.9) NIFAL HIFIL HUFAL· HITPAEL (2 = 0.6%) PAAL· NIFAL· PUAL HITPAEL (1 = 0.3%) PAAL NIFAL PUAL HUFAL (1 = 0.3%) PAAL PUAL HIFIL· HUFAL (1 = 0.3%)

21. NIFAL· PIEL HIFIL· HITPAEL· (1 = 0.3%) 22. NIFAL PUAL· HIFIL· HUFAL (1 = 0.3%)

113

23. FIEL HIFIL HUFAL HITPAEL (l = 0.3%)

Here the only CSS which could occur theoretically, but does not actually, occur is: NIFAL PUAL HUFAL· HITPAEL· That is, a CSS consisting solely of the patterns with valency one, which have been analysed in chapter 4 as resulting from argument reduction. This leads to the conclusion that the NIFAL pattern, which can also be basic (as -tr) cannot serve as input to a derivational rule, for deriving another reduced pattern. In other words, the NIFAL is only basic when it is the only verbal pattern in which a root can occur, or if the other pattern(s) are derived from it by valency increase. The latter alternative, however, does not occur in practice. As will be shown later in this chapter, the NIFAL is derived in the concrete CSS's from the PAAL, PIEL or HIFIL by valency decrease, rather than deriving any of them from the NIFAL by valency increase. Note also that compared with the previous two sub-populations the roots are distributed over the CSS's much more evenly; there is no question of big gaps between one very large CSS against a group of small ones. The two largest CSS's comprise only 24.5% and 15.5% of the subpopulation while the rest is spread out over 21 CSS's, all smaller than 10%; in fact 15 CSS's are smaller than 5% of the sub-population. As we shall see the CSS's consisting of 5 and 6 bioyaaim are distributed similarly. This means that whichever productive processes may be found in these CSS's, they do not hold for a large group of roots (except those involving PIEL, PUAL and HITPAEL in the same functions as in CSS 1 in table (10) and CSS 1 in table ( I D ) . This sub-population consists of the following CSS's (above the limit of (12) CSS 29.1

Bl

B2

B3

B4

29.3

+tr pp -tr i -tr pp/i

30.1

+tr

pp

+tr

30.2

-tr

pp

+tr pp

31.1 31.2

-tr -tr

32.1 32.2 32.3

+tr pp +tr pp -tr i

33

-tr

29.2

B5 c c c

B6 pp pp pp

B7 Examples t.r.m. 'donate' t.r.d. 'bother' l.a.h. 'be tired'

c c c

24.5

y.z.m. 'take initiative' 46 b.x.h. 'cry' 4

50

15.5

-tr

' . y . f . 'tire' y.h.d. 'together'

5 25

30

9.3

-tr rec i

c.v.r. 'collect' r.a.h. 'see* c.r.d. 'hoarse*

18 2 3

23

7.1

20

7.1

18 1

19

5.9

n.h.g. 'behave, do habitually 1 2 n.f.l. 'fall' 15

17

5.3

S.f.x. 'pour' m.g.g. 'melt'

15

4.6

m.h.k. 'erase'

-tr +tr

c/+tr pp +tr pp

34.1 34.2

c +tr

-tr -tr

35.1

+tr

c

pp -tr

35.2

-tr

c

pp -tr

36.1 36.2

+tr pp -tr i

+tr c

nt= 79

pp

c/+tr pp +tr pp

n= 28 1 50

-tr i

f

. n . g . 'enjoy' m . c . ' . 'finish'

14 1

114

Cont. table (12) sCSS 37.1 37.2

B2

Bl +tr -tr

-tr

38

B4

+tr

B6

pp

B7

Examples

1 -tr g.a.h. 'be proud 1 -tr c.v.h. 'scream (sudd)

-tr

+tr pp

+tr

40

B5

c c

+tr (iter) +tr

39.1 -tr 39.2

B3 +tr +tr

n=

nt=

%

2 11

13

4.0

13

4.0

10

3.1

n.b.a. 'prophesy'

c

pp

S.u.t. 'sail*

8

c

pp

h.n.h. 'to park'

2

c

pp

d.g.m. 'model'

8

2.5

pp

S.l.h. 'shed'

6

1.9

p.h.m. 'carbon*

4

1.2

41

-tr

+tr

c

42

-tr

+tr pp

c

Note the following points about table (12): 1. CSS's numbers 38, 40 and 42 feature a HIFAL (B2) without its prototypical source pattern, the PAAL. These cases will be discussed in section 5.5.1.1. There is, however, no CSS where the PUAL occurs without the PIEL, its prototypical source pattern (except in list (iv) in CSS's numbers 18, 19 and 22, which are below the 1% limit). Nor does any CSS feature the HUFAL without its prototypical source pattern, the HIFIL (except for CSS number 19, which is below the 1% limit). 2. According to the literature and the prototypical derivational relations as described in chapter 4, the HITPAEL can be derived from the PAAL· and the PIEL. Indeed, in table (12) no CSS contains a root in the HITPAEL pattern without also featuring a PAAL or PIEL pattern. (In CSS number 17, which is below the 1%, there is a HITPAEL without PAAL· or PIEL.) 3. There are five CSS's, numbers 33, 38, 40, 41 and 42, where the binyanim are consistent in accordance with hypothesis (2) presented in section 5.0 (and in chapter 3 section 3.2 and chapter 4). One of these CSS's (number 33) is medium sized here (6.2% of the sub-population), but small in relation to the CSS's as a whole (table (18)); the other CSS's are all small, 2.5% or less of the sub-population, and very small in relation to all the roots (0.32%, 0.24% and 0.16%). (v)

5 binyanim (total 257 roots, 1%= 3) 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

PAAL NIFAL PIEL PUAL HITPAEL (61 = 23.7%) PAAL· NIFAL HIFIL HUFAL HITPAEL (46 = 17.9%) PAAL NIFAL PIEL· HIFIL HUFAL· (30 = 11.7%) PAAL· NIFAL PIEL PUAL· HIFIL (21 = 8.2%) PAAL PIEL PUAL HIFIL HITPAEL (17 = 6.6%) PAAL PIEL HIFIL HUFAL HITPAEL (17 = 6.6%) PAAL NIFAL· PIEL· HIFIL· HITPAEL (17 = 6.6%) FIEL PUAL HIFIL HUFAL· HITPAEL (15 = 5.9%) PAAL· PIEL PUAL· HIFIL HUFAL (12 = 4.7%) NIFAL PIEL PUAL· HIFIL HITPAEL (7 = 2.7%) NIFAL PIEL PUAL HIFIL HUFAL· (5 = 1.9%)

12. NIFAL PIEL HIFIL· HUFAL HITPAEL (2 = 0.7%) 13. PAAL· NIFAL PUAL HIFIL HUFAL (2 = 0.7%) 14. PAAL NIFAL PIEL HUFAL HITPAEL (1 = 0.4%)

115 14. 15. 16. 17. 18.

PAAL NIFAL FIEL HUFAL HITPAEL (l = 0.4%) PAAL NIFAL PUAL HIFIL HITPAEL (l = 0.4%) PAAL FIEL PUAL HITPAEL HUFAL (l = 0.4%) PAAL· PUAL HIFIL HUFAL· HITPAEL (l = 0.4%) NIFAL FIEL PUAL HUFAL HITPAEL· (l = 0.4)

In this sub-population too, there are probably CSS's which could occur in principle, but do not actually occur. There is, however, no binyanim combination which could have occurred theoretically, but does not actually occur in the sense of coocurring with some other binyanim· Therefore, looking at the possible but non-ocurring CSS's is not interesting. This sub-population consists of the following CSS's (of over the 1% limit):

(13) CSS

Bl

43.1 43.2 43.3 43.4

+tr -tr +tr +tr

B2 PP PP PP PP

44.1 +tr 44.2 -tr

pp -tr

45.1 +tr 45.2 -tr

pp -tr

46.1 +tr 46.2 -tr 46.3 -tr

pp i -tr

47.1 +tr 47.2 -tr 47.3 -tr

B3 +tr +tr +tr +tr

B4 PP PP PP

B5

B6

PP

B7 -tr -tr ref i

Examples nl.m.d. 'learn' 36 h.u.l. 'take place* 16 g.r.d. 'scratch' 1 p.t.h. open, develop' 8

-tr -tr

a.h.v. 'love' n . u . ' . 'move'

20 26

46

'.z.v. 'leave' k.r.n. 'shine*

21 9

30 11.7

d.k.r. 'sting, jab 1 d.'.x. 'extinguish* h.p.z. 'hurry 1

19 1 1

21

8.2

nt=

%

61 23.7

c c

PP PP

+tr +tr

c c

PP PP

+tr pp +tr pp +tr pp

c c c

+tr +tr +tr

c c c

-tr -tr i

n.g.h. 'jostle* g.'.S. 'storm* y.S.r. 'straight '

3 11 3

17

6.6

-tr

p.r.s. 'slice, spread' '.S.r. 'rich*

16 1

17

6.6

15

5.9

12

4.7

pp pp pp

48.1 +tr

pp

+tr

c

48.2 -tr

i

+tr

c

49

+tr

pp

c

pp i/-tr

50.1 -tr 50.2 +tr

+tr +tr

pp pp

c c

pp pp

51.1 51.2

-tr -tr

+tr c

pp pp

c +tr

52

-tr

+tr

pp

c

-tr i pp

'.c.m. 'huge* "y.c.a. 'exit t.l.a. 'patch*

9 3

p.a.r. 'glorify 1 y.t.r. 'leave over 1

5 2

2.7

x.t.r. 'crown'

The size of the CSS's in this table, as in the previous one, is distributed fairly evenly; that is, there are no big gaps between very large vs. very small CSS. The largest CSS (number 43) covers 32.7% of the sub-population of roots occurring in 5 binyanim, and the smallest (number 52) 1.9%. There is nothing remarkable regarding the binyanim combinations which do not occur, such as CSS*s with reduced binyanim only. As there are 5 binyanim in the CSS's, there must be a bivalent one too. The binyanim distributions in this table follow the prototypical derivational relations of valency reduction as described in chapter 4

116

HUFAL) occurs without its prototypical input pattern (PIEL or HIFIL respectively). The NIFAL pattern does occur without a PAAL, in CSS's numbers 51 and 52, but its function is intransitive, not pseudo-passive. Finally, there is one CSS, (number 52) where the patterns have a consistent semantic-syntactic function. (iv)

6 biayaaia (total 185, 1%=2) 1. 2. 3. 4. 5. 6. 7.

PAAL NIFAL· PIEL· PUAL· HIFIL· HITPAEL· (56 = 30.2%) PAAL· NIFAL PIEL HIFIL HUFAL HITPAEL· (39 = 21.1%) PAAL PIEL· PUAL HIFIL HUFAL HITPAEL (33 = 17.8%) PAAL· NIFAL· PIEL PUAL· HIFIL· HUFAL· (28 = 15.1%) NIFAL PIEL· PUAL HIFIL HUFAL HITPAEL· (19 = 10.3%) PAAL NIFAL· PUAL HIFIL HUFAL· HITPAEL· (7 = 3.8%) PAAL NIFAL FIEL PUAL HUFAL HITPAEL (4 = 2.2%)

The only interesting point about this sub-population is that it is the only one (except the one consisting of 7 patterns) where all the CSS's are above the 1% limit. It is also the sub-population featuring the smallest number of CSS's after those with seven patterns and one pattern (in descending order). This sub population consists of the following CSS's:

(14) CSS

Bl

B2

B3

B4

B5

B6

53.1 +tr pp 53.2 +tr i 53.3 -tr -tr

+tr pp +tr pp +tr pp

c c c

54.1 +tr 54.2 -tr

+tr +tr

c c

pp pp

pp -tr

55.1 +tr 55.2 -tr

B7

-tr i -tr

+tr +tr

pp pp

c c

pp pp

Examples

-tr r.c.h. 'want 1 -tr r.'.d. 'tremble' (sudd) -tr -tr

n=

1 s.t.m. 'fill, clog 48 1 k.m.r. 'to round 5 r.v.h. 'ease' 3

56 30.2 39 21.1

r.s.s. 'spray' s.m.n. ' f a t , oil

5 28

33 17.8

b.d.k. 'check' b.t.h. 'trust'

25 3

28 15.1

19 10.3

pp -tr

+tr +tr

pp pp

c c

pp pp

57.1 57.2 57.3

-tr -tr -tr

+tr pp +tr pp +tr pp

c c c

pp pp pp

-tr ref i

y.c.v. 'stable, put 1 15 h.b.a. 'hide' 2 S.r.s. 'roof 2

58.1 +tr 58.2 -tr

pp -tr

+tr pp +tr pp

pp pp

-tr -tr

y.s.d. 'found' p.t.l. 'twist'

59

-tr

pp

pp

-tr

h.l.h. 'be ill1

c

%

16 23

56.1 +tr 56.2 -tr

+tr

nt=

6 1

7

3.8

4

2.2

CSS's 58 and 59 deviate from the prototypical valency reduction process as described in chapter 4 section 4.6.1. CSS 59 features a PUAL without a PIEL, and CSS 58 a HUFAL· without a HIFIL. These cases are discussed in section 5.5.1.2 .

(vii) 7 biayaaim total : 165 roots PAAL NIFAL· PIEL· PUAL HIFIL HUFAL· HITPAEL (165)

Although this sub-population, has only one CSS, the binyanim may have several functions. These are (15)

117

(15) CSS 60.1 60.2 60.3 60.4

Bl B2 +tr pp -tr -tr +tr pp -tr -tr /i 60.5 +tr pp 60.6 +tr pp 60.7 -tr -tr 60.8 +tr ref

B3 +tr +tr +tr c c +tr +tr +tr

B4 pp pp pp pp

B5 c c c +tr

B6 pp pp pp pp

B7 -tr -tr i i

PP pp pp pp

c c c c

PP pp pp pp

i ref ref ref

Examples m.c.a. 'find 1 r.t.h. 'boil 1 n.n.h. 'appoint* p.l.g. 'divide' '.k.m. s.r.k. r.u.m. n.'.r.

n= 64 39 22 24

% 38.8 23.6 13.3 14.5

7 5 3 1

4.2 3.0 1.8 0.6

'twist' 'comb' 'heighten' 'shake off

This is the CSS with the largest number of different binyaaia functions. On average one binyan has 2-3 functions within a CSS. The great variety in this CSS may be due to a. its large size b. its members being very frequently used verbs; The first reason is undermined by the fact that the two other largest CSS's, namely numbers 8 and 16, consisting of 221 and 432 roots respectively, do not exhibit a similar variety in the binyanim functions. CSS number 8 has only two sub-divisions, and CSS number 16 five. It is, therefore, more likely that it is the high frequency of the verbs that creates the great variety in this CSS. It is possible that frequent use leads to a greater diversification in functions. It is interesting to note that the PUAL and HUFAL patterns retain the same function: pseudo-passive throughout, but are not interchangeable. 5.3.1 The size of the CSS1· The CSS's can be further divided into three groups, according to their number of roots. There are large CSS's (of over 100 roots), medium size CSS's (of 50-100 roots) and the small CSS's, of which only those with more than 10 roots are listed here. The % column indicates in percentages the size of the CSS's in relation to the whole population of the 2452 roots. The groups are (only9 the major binyanim functions within the CSS's are listed) the following: (16) Large CSS's CSS Bl B2 17 8 60 +/-tr pp 3

B3 B4 +tr/c pp +tr/c pp +tr/c pp +tr

B5

B6

B7 -tr

c

pp

-tr/i

n= 432 221 165 101

%= 17.61 9.01 6.71 4.11

118

(17) Medium size CSS's: CSS

Bl

B2

9 10 12

+tr

pp

B3 B4 +tr/c

50 51 45

2.07 2.07 1.83

pp

79

3.22

-tr/ref 61 -tr 56

2.48 2.28

pp

c

-tr +tr

19

+/-tr i/pp

c

29

+/-tr pp/i

c

43 53

+tr +tr

pp +tr pp/-tr +tr

pp

c

30

+tr

%= 3.95 3.42 3.22

B6

18

pp

n= 97 84 79

B5

B7 -tr/i

pp

pp pp

c

The small CSS 's (containing less than 50 and more than 10 roots) are

(18) CSS 1 5 7 14 13 15

Bl -tr

B2

B3

B4

B5

B6

n= 85 31 38 25 32 14

%= 3.46 1.26

24 21 19 -tr 19 -tr/i 17 11

0.97 0.85 0.77 0.77 0.69 0.44

-tr -tr

30 23 20 19 17 15 13 13 10

1.22 0.93 0.80 0.77 0.69 0.61 0.53 0.53 0.40

30 21 17 17 15 12

1.22 0.85 0.69

0.69 0.61 0.48

B7

+tr -tr

+tr

+tr -tr -tr +tr

+tr/c -tr

20 21 22 23 24 25

-tr -tr +tr -tr

31 32

-tr +tr

pp/i

33 34 35 36 37 38 39

-tr

-tr

-tr +tr -tr -tr

pp -tr

+tr

+tr +tr

c

pp

c

pp

pp

pp/i

+tr

pp -tr

+tr +t

pp pp

+tr +tr •«•tr +tr

c c c c

pp

c

-tr -tr -tr -tr -tr

1.54 1.01 1.30 0.57

c

pp

c c c c c c

pp

c

pp

-tr

39

1.59

45 46 47 48 49 50

+/-tr pp +tr pp -tr +tr pp -tr

+tr +tr +tr +tr +tr +tr

54

+/-tr pp

+tr

55 56

-tr +tr

+tr +tr

pp pp

c c

pp pp

-tr

33 28

1.34 1.14

+tr

pp

c

pp

-tr

19

0.77

58

pp -tr

pp pp pp pp

pp pp

-tr -tr -tr

119

Note the following points about tables (16)-(18): 1. Only the main functions of the bioyanim are listed in these tables. Thus in table (16) of the 432 roots contained in CSS number 17 only 358 are (+tr) in the FIEL and 347 are (-tr) in the HITPAEL. In CSS number 8 only 212 of the 221 roots are (+tr) in the PIEL; the variation in CSS number 60 is even greater. The same holds for tables (17) and (18). This, however, is of no concern as these tables are meant to present only a general picture of the root distribution. Sub-hypotheses (i) - (ii) regarding the consistency of the binyania's function and valency are not evaluated on the basis of these tables but of tables (10) -(15). 2. The numerical distribution of the roots is more or less equal; 653 roots in the two big CSS's, 655 roots in the medium sized CSS's and 644 roots in the small CSS's (the last group in fact contains more roots, the CSS's with less than 10 roots not being listed). 3. A comparison of the number of binyania with the size of the CSS, shows that the CSS's taking many binyania (5-7) generally contain more small CSS's than those taking less binyania. This is, of course, logical, since the more members a group contains the more combinatorial possibilities it has. 4. The only generalization that can be derived from tables (16)-(18) is that the binyao FIEL (+tr) is productive as it occurs in this function in all the big CSS's; numbers 3, 8, 17 and 60). Binyania FIEL (+tr/c): PUAL (pp) are also productive; they occur in this function in the only two big CSS's as well as in many others. CSS number 17 shows that the HITPAEL pattern is productive in expressing the intransitive. 5.4 Verification of the hypotheses Let us return to the three hypotheses presented at the beginning of the chapter, and check them against the data presented in tables Sub-hypothesis (i) The first sub-hypothesis was that the binyanim have a consistent semantic-syntactic function in the CSS's. That is, whereas throughout the whole system they have two or three different functions, they have only one function within a CSS. This holds in particular for the multi-functional binyania PAAL, NIFAL, FIEL, HIFIL and HITPAEL. (Remember that a CSS consists of all the lines with an identical first digit; a single horizontal line will be referred to as sub-CSS, or sCSS.) The concrete data in tables (10) - (15) show that this hypothesis will not stand up. There are several CSS's in which the multi-functional binyanim also have only one semantic-syntactic function, but these are rare and their number of roots is small. These CSS's are the following:

120

(19) CSS 28

CSS's with consistent bioyanim functions: Bl B2

B3 +tr

B4

33 -tr -tr 38 -tr 40 -tr 41 42 +tr pp 43 -tr

+tr +tr PP +tr PP +tr

50 53

-tr

B5 c

B6 PP

B7

n= 7

c

20

-tr

13

PP PP

PP PP

c c c c

8 6 5 4

+tr PP +tr PP

c c

PP PP

i/-tr 15 5

+tr

Total

82 = 3.32% (of all the roots)

In all the other CSS's the multi-functional bioyanim have two, sometimes three functions. Strikingly, the bioyania in table (19) all feature their main syntactic-semantic functions; the only exception being the FIEL, which is transitive instead of causative (+tr is its secondary function); the HIFIL is causative, not transitive or inchoative; the NIFAL when it occurs without the PAAL is intransitive, inchoative or reflexive; and finally, the HITPAEL is intransitive and not inchoative, reflexive or reciprocal. Note, however, that no serious general conclusions can be drawn from this, because of the limited number of CSS's and their roots. As for the other inconsistent CSS's, note that the greatest diversity occurred in the big and the large medium-sized ones, which is only to be expected. For the more roots a CSS contains, the more possibilities it has to create new combinations. It was alredy pointed out at the beginning of this chapter (in section 5.0) that total consistency cannot be expected.

Sub-bypotbesis (ii) The result with sub-hypothesis (i) can only be fully evaluated in the light of the second sub-hypothesis, namely that the valency of the biayaaim in the CSS's is consistent. Here too, no total consistency should be expected. Since the patterns vary in their syntactic-semantic function, their valency, being partly an expression of that function, will also vary. Nevertheless, tables (10)-(15) show that the quantitative valencies of the patterns are considerably more consistent than their semantic-syntactic functions. The only exception here is the PAAL, which often has valency 1 and valency 2 in the same CSS. For the rest, there are only two CSS's where the same binyan has two different valency values: PIEL, having in one case valency 2 (+tr, c) and valency 1 (-tr), and HIFIL having valency 2 (causative, transitive) and valency 1 (inchoative). It is important to look separately at the valency of a pattern in addition to its semantic-syntactic functions as the valency is not influenced by the construction in which the verb occurs. Whether for example, a verb in the HIFIL pattern is causative or transitive depends partly on the construction (i.e. on arguments), and also, of course, on the concrete root. The function of a verb is sometimes difficult to establish in isolation, whereas a construction (which is in a way a context) makes it clearer. The quantitative valency on the other hand does not depend on the construction. Thus if a verb is bivalent, it will always have two arguments. This is particularly important in Hebrew, because, as shown in chapter 4, the notion of 'transitivity' does not always give an adequate

121

description of the number of (obligatory) arguments a verb in a certain binyan will have. Furthermore, the notion of valency is more suitable for dealing with binyanim in general, i.e. in connection with predicate schemes rather than specific roots. A possible objection could be that since reflexive, reciprocal and inchoative are by definition derived by 2nd argument reduction, it is circular to say that they are univalent. However, the argument here follows a different course. The description of the binyanim functions was first established on the basis of concrete examples; then the other binyaaim (and their functions) were derived fro· the same root. Finally, these facts determined whether a certain binyan from a specific root should be derived by valency operation. Thus the semantic-syntactic functions and the valency entail two different levels of looking at the binyanim. But they are, of course, connected to each other. Therefore, the conclusion that there still is some consistency in the valency of the binyanim on the level of CSS's is not without importance. Sub-bypotbesis

(iii)

The third sub-hypothesis presented at the beginning of this chapter claimed that the same processes of quantitative valency changes apply on the level of the CSS's as on the level of the predicate formation rules. The tables (10)-(15) do not immediately bear this out clearly. There are CSS's in which the same processes clearly do occur, eg. CSS's 8, 9, 10, 17, 24, 29, 32, 40 and 44. Many others, however, do not appear to give a clear picture, eg. CSS's numbers 13, 14, 16, 19, 20, 26 and 37. The next section deals with this sub-hypothesis, and looks mainly at those cases where the derivational processes described in chapter 3 do not seem to apply. Finally at the end of section 5.5, I will return to the evaluation of this sub-hypothesis.

5.5 The derivational relations The relevant derivational relationships within CSS 's can be represented by the following general diagram:

(20)

Bl (+tr) B3 (+tr) B5 (c) "" B7 (-tr) -v | -v| B4 (pp)

B6 (pp)

The predicate-schemes associated with this diagram are the following: PAAL (Bl) (xl)Ag (x2)Go MIFAL (B2) (xl)Proc FIEL (B3) (xl)Ag (x2)Go PUAL (B4) (xl)Proc HIFIL (B5) (xO) Causer (xl)Causee [{x2)Go] HUFAL (B6) (xl)Proc HITPAEL (B7) (xl)Po

([]= optional)

This diagram represents the relations for some of the roots occurring in all 7 bioyaoia, as for example a.x.l. 'eat'. There may be a difference in the number of binyanim or their semantic functions in such diagrams, depending on the CSS. The description of the derivational relations within the CSS 's is based on the following two assumptions:

122

1. It is desirable to retain the same type of derivational processes both for the productive prototypical and the unproductive derivational relationships between the binyanim, so that the whole description is based on the same principles (i.e. valency change). This makes the description simpler, and therefore methodologically better (following the principle of Occam's razor). In chapter 4 most of the productive derivational relations could be described as changes in the quantitative valency of a binyan, where the root + binyan is regarded as a predicate. For example (21)

c.y.r. V-B3 (xl)Ag (x2)Go

'draw V (xl)Ag (x2)Go'

is such a predicate frame. From a methodological point of view, it is desirable to try and keep as far as possible to the same type of processes for describing the non-productive derivational relations as those for describing valency change. This means that for the CSS's too, it is necessary to establish a basic binyan, which will undergo the valency changes. 2. Every CSS contains at least one basic binyan. Since the semanticsyntactic function of some of the binyania is not consistent, more than one type of derivational relationship per CSS is possible. This implies that there can be more than one basic binyan. (i.e. one basic pattern for each derivational network in the CSS). Here too, as in the case of the Stereotypie valency changes described in chapter 4, criteria for establishing the basic binyan are required. The criteria established in chapter 4, which also apply here, are the following: (i) semantically basic (ii) morphological simplicity (iii) consistency with the rest of the system (iv) markedness For a detailed discussion of these criteria see chapter 4, section 4.2. Note that theoretically the criteria may clash. Semantic simplicity, for example, may in a certain case clash with the consistency criterion. In practice, however, this hardly ever happens. The PAAL is always unmarked, and the semantic and morphological simplicity overlap in most cases. The illustration below shows how these criteria may be used in a concrete CSS. For CSS number 42 NIFAL· ( B 2 ) , PIEL· (B3), PUAL (B4) and HIFIL (B5) the following three derivational relationships are possible: (22) a

B3 (+tr/c)

'\

B5 (+tr/c)

B7 (-tr/i)

-v B4 (pp)

The predicate schemes belonging to this diagram are: NIFAL (B2) (xl)Po PIEL (B3) (xl)Ag/causer (x2)Go/causee PUAL (B4) (xl)Proc HIFIL (B5) (xO)causer (xl)causee (x2)Go or: (xl)Ag (x2)Go HITPAEL (B7) (xl)Po

123

(22)b

B3 (+tr/c)

-v B7 (-tr/i)

B5 (+tr/c)

The predicate-schemes belonging to this diagram are: FIEL (B3) (xl)Ag/causer (x2)Go/causee NIFAL (B2) (xl)Po PUAL (B4) (xl)Proc HIFIL (B5) (xO)causer (xl)causee (x2)Go or (xl)Ag (x2)Go HITPAEL (B7) (xl)Po (22)c

B7 (-tr/i)

The predicate-schemes belonging to this diagram are: NIFAL (B2) (xl)Po PIEL (B3) (xl)Ag (x2)Go HIFIL (B5) (xO)causer (xl)causee (x2)Go or (xl)Ag (x2)Go HUFAL (B6) (xl)Proc HITPAEL· (B7) (xl)Po Here (22b) is preferable to (22a) on the basis of morphological simplicity. In choosing between (22b) and (22c) two criteria clash: morphological simplicity would lead to (22b) (because PIEL is morphologically simpler than NIFAL); at the same time (22c) is more consistent with the rest of the system, the HIFIL being derived from the PIEL as in (22b), while in (22c) the HIFIL and the PIEL are both derived from the NIFAL·. In all the other CSS's, the HIFIL as causative is always derived from the PAAL and never from the PIEL. In this case the criterion of consistency with the rest of the system is also strengthened by the fact that it represents the only derivational direction which actually occurs. Therefore, the derivational relationships within this CSS are best represented by (22c). 6 It is important to point out that the analysis of the derivational relations in the rest of this chapter cut across the CSS's, which means that this analysis does not deal with individual complete CSS's but with the functions of the binyania they contain. In the tables illustrating the functions of the binyanim only the relevant parts of the CSS's are quoted. 7

124

5.5.1 Valency reduction -v 5.5.1.1 PAAL > NIFAL· In all CSS's featuring B1-B2 in the prototypical relationship, they can occur together with all the other biayanim, without restricting the functions of those binyanim. The Bl ->B2 in the prototypical relationship of valency reduction occurs in the following types of CSS's: -v

(23)

Bl (+tr)

> B2 (pp): summary of types of CSS's

Bl

B2

B3

+tr

pp

+tr

ref

+tr pp c +tr pp

B4

B5 +tr c

B6 pp

B7 m i ref,

rec,

The derivational relationship here is -v PAAL· (+tr) --> NIFAL· (-tr/ r e f ) . In order to see whether the other binyanim and their functions affect these relations, it is interesting to look at the cases in which the prototypical relation of valency reduction does not apply, i.e. when there is a NIFAL (pp) but no PAAL· (+tr). This happens in the following CSS's (Note that this table only includes the relevant part of a CSS, i.e. the bioyanim functions instead of whole CSS's. The same holds for all comparable tables in this chapter.):

(24)

(-tr) PAAL· --> (-tr) NIFAL·:

sCSS 11.1

Bl

B2 -tr

19.2 19.3 20..2 24..3

-tr -tr -tr -tr

-tr -tr/i -tr -tr

29 3 30.2 32.3 36.2

-tr -tr -tr -tr

i i i i

B3 -tr

B4

B5

B6

B7

-tr

'be cross eyed' 2 'disappear' 41 'finish 1 3 'spit' 2

-tr

t.r.d. b.x.h. m.h.k. S.f.x.

'bother 1 'cry* 'erase' 'pour'

+tr

+tr +tr

PP PP

-tr 43.2 44.2 -tr 45 2 -tr -

PP i -tr

+tr

46.2 48.2

-tr -tr

i i

+tr +tr

PP

c c

53.3 54.2 58.2

-tr -tr -tr

-tr -tr -tr

+tr +tr +tr

PP

PP

PP

c c c

60, 60. 60.7

-tr -tr -tr

-tr -tr/i -tr

+tr c +tr

PP PP PP

c +tr c

pp pp pp

c c c

PP PP PP

n= 2

p.z.l. g.u.z. x.l.h. r.k.k.

+tr c

c c c

Examples a.n.k. 'sigh*

-tr

h.u.l. 'take place* 16 n.u.'. 'move 1 26 k.r.n. 'shine, radiate'19 h.p.z. 'hurry' '.§.r. 'rich 1

-tr

1 4 20 14

r.v.h. 'ease' 1 r.'.d. 'tremble b.t.h. 'trust 1

-tr r.t.h. 'boil 1 i p.l.g. 'divide' ref r.u.m. 'raise 1

1 3 3 23 3 39 7 3

125

In most of these CSS's the NIFAL is derived from the PAAL by qualitative changes of valency, for example of the following type:

(25)

gaz 'disappeared' - nagoz 'disappeared' (+sudden) ra'ad 'shivered' - nir'ad 'shivered ' (+momentaneous)

In some other cases the NIFAL is derived not from the PAAL but from the HIFIL. For example (25)

hexSil 'made f a i l ' - nixSal 'failed' hitrim 'made contribute 1 - nitram 'was donated'

More examples of derivations of the NIFAL from the HIFIL will be given later in this section. In other cases there is no semantic difference between the PAAL and the NIFAL· (eg. sCSS's (=sub-CSS) 11.1, 19.3, 24.3 ). Another deviation from the derivation (+tr) PAAL· —> NIFAL· occurs in CSS's where there is no PAAL at all. These are the following: (Note again, the CSS's are not quoted completely, but only the biayania functions under discussion.) (27) sCSS 21.1 21.2 38

no PAAL: Bl B2 -tr i -

-tr

B3

B4

+tr

pp

B5 c c

B7

Examples m.r.c. 'zeal 1 h.l.d. 'rust'

-tr

n.b.a. 'prophesy 1 13

41

-

-tr

+tr

42

-

-tr

+tr

pp

c

50.1 50.2 52

-

-tr -tr -tr

+tr c +tr

pp pp pp

c +tr c pp

57.1 57.2

-

-tr -tr

+tr +tr

pp pp

c c

pp pp

-tr ref

y.c.v. 'stable' h.b.a. 'hide'

57.3

-

-tr

+tr

PP

c

pp

i

S.r.S. 'root'

These CSS's PIEL or HIFIL) the HIFIL·, not resultative. 8 (28)

c

B6 pp pp

pp

S.l.h. 'shed '

p.h.m. 'carbon 1 -tr i i

n= 19 1

6

4

p.a.r. 'glorify* 5 y.t.r. 'leave over 2 x.t.r. 'crown* 5 15 2 2

always feature a transitive or causative pattern, (i.e. from which the NIFAL is derived. Note, however, that it is the PIEL. The derivational relation is causative —> For example

heyciv 'put, made stable' —> hehbi 'made hide' —>

nicav 'stood 1 (from y.c.v.) nebba 'hid' (from h.b.a.)

Generalizat ions: 1. A verb in the NIFAL pattern can only be active intransitive or inchoative when the PAAL is -tr or when there is no PAAL in the CSS. 2. In such cases the NIFAL is mostly not pseudo-passive but active-intransitive or inchoative (resultative). The inchoative function of the NIFAL· is illustrated below: (For the discussion of the difference between -tr, pp. i, ref and rec see sections 4.6.2. and 4.6.3.) (29)a

hamenuha hirgi'a (B5) oto. the-rest made-calm-3p.s.f. him 'The rest calmed him down'.

126

b

hu nirg'a (B2) ketocaa mehamenuha. he calmed down as-result-of the-rest 'he got calm as a result of resting/the rest 1 . hahadasot hir'iau (B5) et kulam. the-news made-excited GM everyone "The news excited everyone 1 , kulam nir'aSu (B2) mehahadaSot. everyone got-excited from-the-news 'Everyone became excited from the news'.

(30)a b

The valency relation between the NIFAL and the transitive pattern in the CSS i.e. FIEL or HIFIL can be seen as either valency reduction of the PIEL or HIFIL, or as valency increase of the NIFAL. That is, it can be either -v

(31)a

-v

PIEL·

> NIFAL· ; HIFIL·

> NIFAL·

or +v

(31)b

+v

NIFAL·

> PIEL· ;

NIFAL·

> HIFIL·

If the syntactic function of the NIFAL· in these CSS's were pseudopassive, it would be derived from the HIFIL or the PIEL by valency reduction (of the type of first argument reduction)in order to maintain consistency with the rest of the system. But according to table (27) the NIFAL in CSS's without a PAAL (i.e. with PIEL or HIFIL as the only bivalent pattern) is intransitive or inchoative. Therefore, the direction of the derivation must be determined by other criteria. Of the two alternatives in (31) the first, valency reduction of the PIEL and the HIFIL, is preferable because it is morphologically simpler than the NIFAL. So we get -v

PIEL·

-v

> NIFAL· ;

HIFIL·

>

NIFAL·

The third type of deviation from the prototypical valency reduction PAAL (+tr) —> NIFAL· (pp) occurs when there is a PAAL· (+tr), but no NIFAL. The CSS's in question are the following: (32) sCSS 13.2 14.1 15.1

CSS's without B2 Bl B2 B3 64 +tr +tr +tr +tr -

18.2 22.2 25.2 26.2

+tr +tr +tr +tr

-

35.1 37.1 39.2

+tr +tr +tr

47.1 55.1

B5

B7

c -tr c

+tr

B6

pp

c +tr

-

+tr +tr

c c c

+tr

-

+tr

pp

c

+tr

-

+tr

pp

c

pp

-tr -tr -tr

n.h.g. 'behave, habit 1 2 g.a.h. 'be proud 1 2 h.n.h. 'to park 1 2

-tr

n.g.h. 'jostle'

3

-tr

r.s.s. 'spray 1

5

pp

pp

'pleasant 1 'furnish 'limp' 'burp'

n= 3 23 1

n.'.m. r.h.t. h.g.r. g.h.k.

pp

+tr

Examples m.r.h. 'disobey' s.b.r. 'break' a.v.h. 'want, wish'

2 15 1 2

127

Note that this deviation from the prototypical derivational pair does not carry any further consequences for the system, valency reduction being an optional process and not an obligatory one. Therefore, it does not have to occur in all cases where it theoretically could. Nor has such a group of verbs that does not undergo valency reduction any influence on the analysis of the other cases. It is perhaps interesting to note that when there is no PAAL the NIFAL can in many cases be derived from the HIFIL, but when there is no NIFAL, there are no PUAL or HUFAL patterns derived from the PAAL. This means that the possibility of getting a pattern other than the prototypical one to fill the pair is not symmetrical. The reason for this might be that the NIFAL can be multi-functional and when derived from the HIFIL it is in fact not pseudo-passive but inchoative-resultative. Such a relationship is not possible between the PAAL and the PUAL/HUFAL. -V

5.5.1.2 PIEL

-v

> PÜAL; HIFIL·

> HUFAL·

The other prototypical valency reduction of the first argument reduction type occurs between the PIEL - PUAL· and the HIFIL - HUFAL·, where the PUAL· and HUFAL are the pseudo-passive of the PIEL and HIFIL respectively. This valency reduction relation holds for all cases featuring both PIEL - PUAL and HIFIL - HUFAL. The HUFAL can result from double valency reduction too, from the HIFIL as described in chapter 4 section 4.6.4 (rule 98). The PUAL and the HUFAL have no other function than the pseudo-passive, they always result from valency reduction. Let us now look at the CSS's which deviate. In the following CSS's the prototypical input predicate PIEL (B3) or HIFIL (B5) is missing: (33)

no PIEL

sCSS 59.1 59.2

Bl B2 +tr pp +tr -tr

(34)

no HIFIL·

sCSS 59.1 59.2

Bl +tr -tr

B2 pp -tr

B3 -

B4 pp pp

B5 c c

B3 B4 +tr pp +tr pp

B5 -

B6

B7

pp

-tr

B6 pp pp

B7 -tr -tr

Examples h.l.h. 'be ill 1 s.m.k. 'blush'

Examples y.s.d. 'found' p.t.l. 'twist 1

n= 3 1

n= 6 1

In these CSS's the PUAL· and HUFAL always occur with another transitive pattern. In some cases this other transitive biayan is the PIEL or the HIFIL. In other cases there are two patterns that are transitive, the PAAL· and the PIEL/HIFIL·. 9 The question then is whether the PUAL· or HUFAL· are to be derived from the PAAL or from the PIEL/HIFIL. In other words, is it (35a) or (35b):

(35)a

Bl (+tr) -v B4(pp)/B6(pp)

or

B3(+tr), B5(+tr)

128

(35)b

-v

Bl(+tr) —> B3(+tr)

> B6(pp) -v

Bl(+tr)

> B5(+tr)

> B4(pp)

The predicate-schemes belonging to these patterns here are: PAAL (Bl) (xl)Ag (x2)6o FIEL (B3) (xl)Ag (x2)Go PUAL (B4)(xl)Porc HIFIL (B5) (xl)Ag (x2)Go HUFAL (B6) (xl)Proc There are no semantic arguments to answer the question of the the choce between (35a) or (35b), in that the PUAL· or HUFAL might be derived from either the PAAL/HIFIL· or PAAL/PIEL respectively. In the following example intuitively the HUFAL seems to be related to the PAAL (as in 36a) rather than to the PIEL, that is, if the 'active' form is with a verb in the PAAL, its pseudo-passive equivalent would be in the NIFAL form, and not in the PUAL or HUFAL·; but if the active predication is in the PIEL form (as in 37a), its pseudo-passive equivalent is in the PUAL or HUFAL pattern, not the NIFAL. This meaning relation, however, holds only for the example below (and perhaps some more cases); I cannot give any general tendency or rule regarding the other cases. (36)a

hem yasdu (Bl) et hakeren hazot lifney 5 äanim. they founded GH the-foundation the-this ago 5 years 'They founded this foundation 5 years ago 1 , hakeren hazot nosda (B2) lifney 5 Sanim. the-foundation the-this was-founded ago 5 years 'This foundation was founded 5 years ago. 1 hem yisdu (B3) et hakeren hazot lifney 5 sanim. they founded GM the-foundation the-this ago 5 years 'They founded this foundation 5 years ago', hakeren yusda (B4) /husda (B6) lifney 5 äanim. the-foundation was-founded ago 5 years 'The foundation was founded 5 years ago.'

b (37)a b

There is also a non-prototypical relation in CSS's featuring a PIEL· or HIFIL, or both, but no PUAL or HUFAL· respectively. Note that the valency reduction for deriving PUAL and HUFAL· is optional. (38) sCSS 9.1 9.2 13.1 13.2 16.1

no PUAL Bl B2

20.1 20.2 23.1 23.2 26.1 26.2 26.3 28

+tr -tr -tr +tr -tr +tr -tr

-tr +tr pp -tr

B3 +tr c +tr +tr +tr

B4 -

+tr +tr +tr +tr +tr +tr +tr +tr

-

B5

B6

B7 -tr i

c

-tr -tr c +tr c c pp

Example s.m.l. 'symbolize' s.h.r.r. 'free' g.v.'. 'die' m.r.h. 'disobey' h.p.a. 'defend' c.r.m. x.l.h. h.l.x. n.c.c. s.h.k. g.h.k. r.c.d. d.b.r.

n= 80 17 6 3 2

'disturb 1 21 'finish 1 3 'walk' 17 'shine' 2 'discolour 1 6 'burp' 1 2 'glitter 1 'destroy insects' 7

129

Cont. of table (38)

B3 +tr +tr c +tr +tr

B4 -

B5 B6 c

-

+tr

+tr +tr

-

c c c c

45.1 45.2 48.1 48.2

+tr PP +tr PP +tr PP -tr i

+tr +tr +tr +tr

-

c PP +tr PP c c

54.1 54.2

+tr -tr

+tr

-

c

•t-tr

sCSS 33

36.1 36.2 37.1 37.2 39.1 39.2

B2 -tr PP i

Bl -tr +tr -tr +tr +tr -tr

PP -tr

B7 -tr

-tr -tr PP PP

-tr

i

n=

' . Z . V . ' leave' k.r.n. ' shine* p.r.s. ' slice1 '.s.r. ' rich'

21 19 16 1

20 14 1 2 11 8 10

-tr r.c.h. ' want' -tr r . ' . d . · tremble'

PP PP

c

Examples m.h.k. ' erase' S.f.x. ' pour* m.g.g. · melt' g.a.h. ' be proud' c.v.h. ' scream' s.u.t. ' sail1 h.n.h. ' to park'

16 23

(sudd) Here in some cases the PIEL can undergo valency reduction. The output biayao is the NIFAL. This is possible for example in the following cases:

(39)

FIEL Siher mogeg cimek yiter

'wished 1 'made disappear' 'made shrink* 'left over'

-

NIFAL· niShar 'was wished' namog 'has disappeared' nicmak 'shrunk, got shrunken 1 notar 'got/was left over'

In fact here the PIEL expresses a certain degree of causativity, and the NIFAL expresses inchoation (resultative). The NIFAL in such pairs is never pseudo-passive (the latter being derived only, from the PAAL). The other group of roots also deviating from the prototypical derivational relation owing to the lack of the reduced pattern is formed by the roots that occur in the HIFIL, but not in the HUFAL pattern. The following CSS's are involved: (40) no HUFAL sCSS 13.1 13.2 13.3 13.4

Bl B2 -tr +tr -tr i

19.1 19.2 19.3 25.1 25.2 26.1 26.2 26.3 32.1 32.2 32.3

+tr -tr -tr -tr +tr -tr +tr -tr +tr +tr -tr

B3

PP -tr -tr +tr +tr +tr PP PP i

B4

B5

B6

+tr +tr c

-

c/i

-

c c c c -»-tr +tr c c c

B7

-tr -tr

-tr

rec i

Examples 1 g.v.'. * die m.r.h. ' disobey1 n.d.m. * quiet ' c.h.v. ' yellow* k.d.h. p.z.l. g.u.z. g.s.s. h.g.r. S.h.k. g.h.k. r.c.d. c.v.r. r.a.h. c.r.d.

n= 6 3 22 1

2 ' dig' * look cross eyed' 2 41 ' disappear ' 10 ' be coarse* * limp' 1 ' discolour ' 6 ' burp' 2 1 ' glitter 1 18 ' collect' 2 ' see' ' be hoarse* 3

130

Cont. of table (40): sCSS 33 34. 1 34. 2 37. 1 37. 2 42 46. 1 46. 2 46. 3 47. 1 47. 2

B2 B3 -tr +tr c/+tr +tr +tr +tr -tr +tr PP +tr i +tr -tr +tr +tr +tr

Bl -tr +tr -tr +tr -tr -tr +tr -tr

47. 3 -tr 48. 1 +tr 48. 2 -tr 51. 1 51. 2

+tr +tr PP i +tr -tr +tr -tr c

B7

Examples m.h.k. 'erase 1

-tr -tr -tr -tr

-tr -tr

'.n.g. m.c.'. g.a.h. c.v.h. p.h.m. d.k.r. d.'.x. h.p.z. n.g.h. g.'.s.

n= 20 'enjoy' 18 'exhaust, finish' 1 'be proud' 2 11 'scream' 'carbon 1 4 'jab' 19 1 'extinguish' 1 'hurry' 'jostle' 3 'storm' 11

c c

c PP c PP +tr -

i -tr i -tr i

y.s.r. p.r.s. '.S.r. p.a.r. y.t.r.

'straight* 'slice, spread' 'rich' 'glorify 1 'leave over'

pp c pp c pp c

-tr i -tr

s.t.m. 'fill, clog' k.m.r. 'round' r.v.h. 'ease 1

B4 PP PP PP PP PP PP PP PP PP

53.1 +tr pp +tr 53.2 +tr i +tr 53.3 -tr -tr +tr

B5 c c +tr c c c c c c c c

B6

3 16 1 5 2 48 5 3

The conclusion from table (40) is that the HIFIL is mostly causative and not transitive in CSS with HIFIL· (c) and PAAL (-tr) and NIFAL (-tr) as in: (41) HIFIL PAAL· hegiz 'made disappear' gaz 'disappeared 1 hehpic 'made desire' bafac 'desired'

NIFAL

root

nagoz'disappeared' g.u.z nehpac 'was desired* h.p.c.

In some cases the NIFAL expresses pseudo-passive, and in others the inchoative-resultative of the action expressed by the HIFIL of that root. Here it is impossible to derive the NIFAL· from the PAAL anyway, because the PAAL· is intransitive. The derivation of the NIFAL· from the HIFIL· is also supported by the semantic relation between an action expressed in the HIFIL· and its result expressed in the NIFAL. Thus PAAL· is the unmarked binyan, but on semantic grounds NIFAL is not derived from it but from the HIFIL·: +v

PAAL· b. (42)

-v

>

HIFIL·

>

NIFAL·

CSS's with HIFIL·(c) and PAAL(-tr) and HITPAEL(i/-tr) as illustrated below: Bl

B5

B7

root

saag 'roared*

hisig 'made roar'

Sana 'stayed'

hisha 'delayed'

histaeg 'roared suddenly' s.a.g. (+moment.) histaha 'lingered,stayed' S.h.h.

The HITPAEL, if derived from the HIFIL, results from 2nd argument reduction. It thus has a different status from the NIFAL mentioned above which resulted from first argument reduction. This type of derivation will be dealt with in the next section. 2. Another variation is formed by CSS*s containing the NIFAL· and PUAL, but where they are derived from the PAAL and PIEL respectively, instead

131 of from the HIFIL (for example in CSS number 30). Since CSS's with PAAL as intransitive are exceptional the NIFAL cannot be derived from it. For example: (43)

Bl

B2

B5

root

laa (-tr) 'tired 1 carad (-tr) 'got hoarse 1

nila (-tr/i) 'got tired' nicrad (i, res) 'became hoarse'

hila (c) 'made tired' hicrid (c) 'made hoarse'

l.a.h. c.r.d.

Here too the NIFAL· expresses the resultative of the causative HIFIL, and as a result can be seen as being derived from the HIFIL on semantic grounds. We get: -v (44) HIFIL· (+tr/c) —> NIFAL· (inchoative/resultative) Summing up, it can be concluded that the PIEL· and HIFIL patterns can undergo valency reduction of the following types: a. the same as the prototypical derivations: -v (i) PIEL· —> PUAL· : +tr —> pp (1st argument reduction) -v (ii) HIFIL· —> HUFAL· : +tr/c —> pp (1st argument reduction) b. other derivational relations: -v (iii) PIEL· > NIFAL· : +tr/c —> inchoative (1st argument reduction) -v (iv) HIFIL· —> NIFAL· : c/+tr —> inchoative (1st argument reduction) -v (v) HIFIL· > PUAL· : c/+tr —> inchoative (resultative) -v (vi) PIEL· —> HUFAL· : c/+tr —> inchoative (resultative) As far as I know both the NIFAL· and the HUFAL· are seen in the literature as passives of the HIFIL. I have found no mention of the NIFAL standing in an inchoative (resultative) relation to the HIFIL·· 10

-v 5.5.1.3 PAAL / FIEL / HIFIL·

> HITPASL

General The last binyaa which is derived in the productive cases by means of valency reduction is the HITPAEL. Unlike the NIFAL, PUAL· and HUFAL, the HITPAEL is derived by second argument reduction. As mentioned already in chapter 1, the HITPAEL can have the following semantic-syntactic values: intransitive, inchoative, reflexive and reciprocal; the order of listing represents in descending order the frequency with which the HITPAEL· performs a given function. The majority of the CSS's featuring the HITPAEL contain a transitive binyan which can be the input predicate to a detransitivization rule. The exceptions are: (45) CSS's with B7 and without any ( + t r ) pattern: Bl

B2

B3

B4

-tr -tr

-tr

B5

B6

B7

-tr -tr

Examples

n=

g.u.r. 'live, inhabit 1 14 h.l.h.l. 'dribble, shiver' 5

132

In the CSS's with a -tr PAAL this PAAL is the basic pattern. For example the verb: g.u.r. 'live, inhabit'. In this case the PAAL is the basic pattern in the CSS owing to the following criteria: a. it is morphologically simpler than the HITPAEL; b. the HITPAEL of this root is more literary, and as such marked. Although there is no difference in quantitative valency between the PAAL and the HITPAEL·, there is a slight semantic difference between them. Both gar and hitgorer mean 'live, inhabit 1 , but the latter also implies continuity over a period of time. The PAAL is the more frequent, unmarked form. The other group, which consists of roots occuring in the FIEL (-tr) PUAL (-tr) and HITPAEL (-tr) the HITPAEL· has an additional semantic feature to the FIEL, namely +sudden, or +momentaneous. Therefore, besides being morphologically simpler, the FIEL is also semantically simpler than the HITPAEL·, and is therefore the basic pattern in the CSS. All the other CSS's contain a binyan with two or more arguments, which can, in principle, undergo argument reduction, and thus be input to a rule deriving HITPAEL. In some of the CSS's the only transitive binyan is a causative FIEL or HIFIL. This FIEL or HIFIL is the basic binyao being both semantically basic and morphologically simpler. In the following examples the PIEL and the HIFIL are causative, and the HITPAEL, derived from them, is resultative: (46)a b (47)a b

habinyanim hagvohim hix'iru (B5) et hanof. the-buildings the-tall made-ugly GM the-view 'The tall buildings spoiled the view' ketocaa mehabniya hanof hitka'er (B7). as-result-of from-the-building the-view got-ugly 1 'As a result of the building the view got spoiled hu hi£va (B5) et namehirim behanuyot sonot. he compared GM the-prices in-shops different-pi 'He compared the prices in various shops' hem histavu (B7) 'al hamebir. they reached agreement on the-price 'They agreed on the price*.

Note that in (47a,b) the HITPAEL· is not related directly to the HIFIL·; biStavu means higi'u leSivyon 'reached equality', in this case: 'reach equality of opinions', i.e. 'agree'. Let us proceed to examine the CSS's in which the HITPAEL occurs in the functions intransitive, inchoative, reflexive and reciprocal. (i) Intransitive Since intransitive is the most frequent syntactic function of the HITPAEL, it occurs in most CSS's; it is, therefore, clearer to give a summary table, rather than list all the CSS's separately. Most likely, here as in the PAAL· - NIFAL relation, the exceptions, or the minority CSS's, will be the more revealing. (48)

HITPAEL as intransitive (summary of CSS types)

Bl

B2

B3

B4

B5

B6

B7

-tr +tr

-tr pp

+tr c

pp pp

c +tr

pp pp

-tr -tr

133

The HITPAEL as intransitive can occur in the same CSS together with any other binyaa in any of its possible functions; the NIFAL as inchoative is an exception to some extent. It usually correlates with HITPAEL· when it is also inchoative rather than when it is as intransitive (cf. CSS'S 24, 32, 36, 49, 54). The following examples illustrate the derivational relation of second argument reduction by which the HITPAEL is derived. Example (49) illustrates the change of meaning in the HITPEAL as compared to the PAAL (though here there is no argument reduction). Examples (50) and (51) illustrate the argument reduction; the change of meaning is slighter. In (50) the PIEL refers to a situation which is completed at the moment when the sentence is said, whereas the HITPAEL refers to a process while occurring. The same holds for (51). (49)a

ha'ec n'a (Bl) baruab. the-tree moved in-the-wind 'The tree moved in the wind' ha'ec hitno'e'a (B7) baruah. the-tree moved-iterative in-the-wind 'The tree moved in the wind' ... haaisrad piteah (B3) toxnit mitar lekol haezor. the-office developed plan infrastructure to-whole the-area 'The office developed a plan for the infrastructure of the whole area', ha'ir mitpatabat (B7) bemehirut raba. the-town develops in-speed great 'The town develops very fast*. al tacivu (B5) et hasulhart baknisa, ze mafri'a. not make-stand-you GM the-table in-the-entrance, it disturbs 'Don't put the table in the entrance, it is in the way', bemeSex Sana leabar SeSubrar mehakele, hu haya during year after that-freed from-the-prison he was mityacev (B7) kol Savu'a bamiStara. present-himself every week in-the-police 'For a whole year after being freed from prison he used to present himself every week at the police*.

b (50)a

b (51)a b

The cases when there is no bivalent binyan in the CSS have been dealt with above ( c f . table (45) and its discussion). (ii) Inchoative The inchoative occurs less frequently as a function of the HITPAEL than the intransitive. The inchoative HITPAEL too is derived by 2nd argument reduction. The HITPAEL has the function inchoative in the following CSS's: (52) sCSS

HITPAEL as inchoative Bl

B2

B3

9.2

c

17.3

c

24.3 31.1 32.3

+tr -tr -tr

B4

B5

pp

i +tr/c pp i

c

B6

B7

Examples

i

s.b.r.r. 'dizzy'

17

n=

i

a.x.l.s. 'populate*

69

i i i

r.t.m. '.y.f. c.r.d.

'put in harness 1 3 'tire' 5 'be hoarse* 3

134

Cont. table (52): sCSS 36.2

Bl -tr

B2 i

B3 c

B4

B5

44.4 47.3 48.2 51.2

+tr -tr -tr

pp

pp pp

i -tr

+tr +tr +tr c

pp

c c +tr

53.2 57.3

+tr

i -tr

+tr +tr

pp PP

c c

60.1

+tr

pp

+tr

PP

60.4

-tr

-tr/i

c

60.5

+tr

pp

c

B6

B7 i

Examples m.g.g. 'smelt 1

n= 1

i i i i

p.t.h. y.5.r. '.s.r. y.t.r.

PP

i i

k.m.r. 'to round' a.r.S. 'root 1

5 2

c

pp

i

m.n.h. 'appoint'

22

pp

+tr

pp

i

p.l.g. 'divide 1

24

PP

c

PP

i

'.k.m. 'twist'

Open 1 'straight 1 'rich' 'leave over'

8 3 1 2

7

The HITPAEL inchoative occurs mostly in CSS's featuring a causative biayan; in most cases this causative binyao serves as an input for the rule deriving the HITPAEL. The HITPAEL in this function mostly has an inchoative resultative meaning; i.e. it refers to a State type state of affairs which results from an action or process with the same root in PIEL11 or HIFIL. The derivation is based on the semantic relation cause-> result. For example: (53)a b

(54)a b (55)a b

ha'asakim babursa he'eSiru (B5) oto. the-business in-stock-exchange made-rich him 'The business in the stock-exchange made him rich 1 hu hit'aSer (B7) ketocaa meha'asakim he got-rich as-result-of the-business babursa. in-the-stock-exchange 'He became rich as a result of the business in the stock exchange' hamenubha hizka (B3) oto. the-rest made-strong him 'The rest strengthened him' hu hitbazek (B7) ketocaa mehamenuha. he got-strong as-result-of the-rest 'As a result of resting he got strong* habom yibeS (B3) et hakvisa bemehirut. the-heat dried GH the-laundry quickly 'The heat dried the laundry quickly' ketocaa mehahom/ biglal hahom hakvisa as-result-of the-heat/ because-of the-heat the-laundry hityabSa (B7) bemehirut got-dry quickly 'Because of/ as a result of the heat the laundry dried quickly'.

However, there are also CSS's of this type where no such direct derivational connection between the HITPAEL and a causative pattern is possible. For example

135

(56)a b

hu tiken ve-hizek (B3) et hagader. he fixed and-strengthened GM the-fence 'He fixed the fence and strengthened it' *hagader hitbazka (B7) ketocaa mehatikun the-fence got-strong as-result-of the-fixing

only a lexical expression is possible (56)c

ketocaa mehatikun hagader (na'asta) hazaka yoter. as-result-of the-fixing the-fence (became) strong more 'Due to the reparation the fence became stronger'

A possible reason why the inchoative HITPAEL cannot always be derived from a transitive or causative binyaa within the same CSS is that the inchoative and intransitive HITPAEL is often used in Agentless or generic constructions, and these are not directly related to any activetransitive construction, at least not in a way that they can be paraphrased by a construction using the PAAL or PIEL, because these need an explicit Agent. (It is possible to get an impersonal construction with the PAAL or PIEL with the generic Agent bea 'they, but none of the sentences below can be paraphrased by such a construction either). In the examples (57) - (59) below the predication refers to a situation, whose Source or Cause cannot be pinpointed: (57)

(58)

(59)

leahar miken histaber (B7) Seyosi bixlal lo hitkaven afterwards became-clear that-Yosi at-all not intended lavo to-come 'Afterwards it became clear that Yosi did not intend to come at all.' pSa'im mehasug haze mityaSnim (B7) crimes of-the-type the-this get expired leahar 30 Sana. after 30 years 'This type of crime expires after 30 years' babom haze hakvisa mityabeäet (B7) bekalut. in-the-heat the-this the-laundry gets-dry easily 'In this heat the laundry dries easily*.

Interesting cases are those CSS's where there are two inchoative patterns: the HITPAEL· and the NIFAL. Several examples are (60)

(61)

(62)

ketocaa mikol hahavayot hameSutafot hakeser as-result of-all the-experiences the-common the-tie beyneyhem nehedak(B2) /hithadek(B7) /na'asa haduk. between-them got-strong got-strong was-made strong 'As a result of all the common experiences the tie between them became strong'. pitom nigla (B2) / hitgala (B7) le'eynenu nof suddenly got-revealed/ got-revealed to-eyes-our landscape marhiv. beautiful 'Suddenly a beautiful landscape was revealed to our eyes'. habor nimla (B2)/ hitmala (B7) mey geSem. the-hole got-full / got-full water-of rain 'The hole became filled with rain water.'

136

In most cases there is no semantic difference between the NIFAL and HITPAEL. Sometimes, however, there is a slight difference of meaning: the NIFAL focuses more on the result, the HITPAEL· more on the process, (e.g. with aicrad/ hictared 'became hoarse' or nigla/ hitgala 'got revealed 1 ) There are, however, also many roots where the inchoative can be expressed either by the NIFAL or by the HITPAEL, but not by both. In the following examples in (a) only the NIFAL can be used to express inchoative-resultative, and in (b) only the HITPAEL: (63)a b

neherad 'got scared' - *hithared nifgam 'got damaged' - *hitpagem *ne'e5ar - hit'aSer 'became rich* *nisbar - histaber 'became clear'

There might be pragmatic differences between the usage of two patterns but these are not discussed here. Summing up, the following can be said about the HITPAEL-inchoative: the HITPAEL can be derived from a transitive or causative PlEL or HIFIL. (ii) in some of the CSS's, where both the NIFAL and the HITPAEL are inchoative, they can be used interchangeably.

(i)

(iii) Reflexive As pointed out in chapter 4 section 4.6.3 reflexivity in MH is expressed not only through the HITPAEL, but also pronominally. In fact, the HITPAEL· is not often used to express this syntactic function. It occurs in the following CSS's: (64) sCSS

17.4 42.4 60.6 60.7 60.8

HITPAEL reflexive Bl

+tr +tr -tr +tr

B2

B3

B4

pp pp -tr ref

+tr +tr +tr +tr +tr

pp pp pp pp pp

B5

c +tr c

B6

B7

pp pp pp

ref ref ref ref ref

Examples

n=

S.h.r.r. 'free 1 14 g.r.d. 'scratch 1 1 s.r.k. 'comb' 5 r.u.m. 'raise' 3 n . ' . r . 'shake off 1

The reflexive RITPAEL occurs in CSS's with at least one transitive binyao, which is either -»-transitive or causative, from which it can be derived by second argument reduction. If there is a PAAL (+tr) then the HITPAEL (ref) is derived from it as in CSS's 42.4, 60.6 and 60.8. In CSS 17.4 the only bivalent pattern is the PIEL, which is therefore the input to the derivational rule. In CSS 60.7 there are two possible input patterns: romem (B3) and berim (B5) 'raised'. Both are transitive, and the HITPAEL could therefore be derived from either of them. In all the other CSS's the PAAL or PIEL is also the basic pattern of the CSS, and so deriving the HITPAEL from them is consistent with the derivations so far. In the last case (sCSS 60.7), however, there is a problem. Should the PAAL· (-tr) be considered as the input pattern? If so, why not derive the HITPAEL from it by a process of qualitative valency change, rather than by quantitative valency change, i.e. reduction of the PIEL or HIFIL? I prefer to derive the HITPAEL· from the PIEL· or HIFIL by valency reduction rather than from the PAAL by qualitative valency change for the following reason: semantically the HITPAEL is more closely related to a transitive action than to an

137

intransitive one, for reflexivity implies transitivity (only, the Goal of the action is identical to the Agent). The question whether the HITPAEL of the roots under discussion should be derived from the PIEL· or from the HIFIL must remain unanswered here. In principle, the HITPAEL can be derived either from a causative or from a transitive pattern. In the following examples the reflexive HITPAEL is derived from a causative PIEL or HIFIL: (65)a b (66)a

b

yosi cihceah (B3) et hana'alayim. Yosi polished GN the-shoes 'Yosi polished the shoes', Yosi 'hictahceah' (B7) likrat hamesiba. Yosi preened-himself-up for the-party 'Yosi preened himself up for the party 1 . madu'a lo hivdalta (Bl) bamaamar beyn why not distinguished-2p.m. in-the-article between hanimukim haSonim? the-arguments the-different 'Why didn't you distinguish in the article between the different arguments?' metivo hu tamid haya mitbadel (B7). from-nature-his he always was separate-himself 'By nature he was always on his own*.

The reflexive HITPAEL can be derived also from an active-transitive PIEL or PAAL, as illustrated below. There are, however, no examples of the HITPAEL being derived from the HIFIL. (67)a b (68)a

of a. b. c. d.

hasar 'idxen (B3) et ha'itonaim be'inyan ... the-minister updated GM the-reporters in-matter-of... 'The minister updated the reporters in the matter o f . . . ' hi mit'adkenet (B7) beofen kavu'a be'inyan ha... she updates-reflexive regularly in-matter-of the... 'She updates herself regularly in the matter o f . . . ' ima tamid rohecet (Bl) et hatinok, aval dani mother always washes GN the-baby, but Danny Sehu kvar yeled gadol, mitrahec (B7) levad. that-he already boy big, washes-reflexive alone 'Mother always washes the baby, but Danny, who is a big boy, washes himself alone.'

The derivation of the HITPAEL, can be cases: only possible source for derivation : possible sources for derivation; PAAL possible sources for derivation: PAAL· possible sources for derivation: PAAL HIFIL· (+tr).

summed up in the following types PIEL (c); (+tr) or HIFIL· (c); (+tr), PIEL(-t-tr) or HIFIL· (+tr) ; ( - t r ) , PIEL (+tr/c) or

In all cases the HITPAEL (ref) is derived by valency reduction. (See also 4.6.3.) Case (a) poses no problem: the PIEL pattern is the basic biayaa in the CSS and the source for deriving the HITPAEL. In case (b) the basic pattern of the CSS is the unmarked PAAL. Since this PAAL is (+tr) it can also serve as the source for deriving the reflexive HITPAEL.

138

In case ( c ) , too, the PAAL is the basic pattern. It is unmarked vs. the PIEL, and morphologically simpler than the HIFIL. Since this PAAL is ( + t r ) , it can undergo valency reduction. In case (d) the HITPAEL is not derived from the basic pattern of the CSS. The basic pattern is the PAAL· (being morphologically just as simple as the PIEL and unmarked). But since this PAAL is intransitive, i.e. has only a first argument, it cannot undergo second argument reduction in order to derive the reflexive HITPAEL; the latter is, therefore, derived from either the PIEL· (+tr) or the HIFIL (+tr). Analysis of the actual roots in this CSS as well as other roots, did not reveal any instance in which the reflexive HITPAEL· is related to the HIFIL·. For that reason, it will be derived from the PIEL. Schematically we get the following derivational processes in the CSS's listed in (69):

(69)a /

^

-v (2nd arg) B7 (-tr)

The predicate-schemes belonging to this diagram are: PIEL (B3) (xl)Causer/Ag (x2)Go PUAL (B4) (xl)Proc HITPAEL (B7) (xl)Po

(69)b +v B5 (c)

-v (2nd arg) •\ B7 (ref)

B2 (i)

The predicate-schemes belonging to this diagram are: PAAL (Bl) (xl)Ag (x2)Go NIFAL (B2) (xl)Po HIFIL (B5) (xO)causer (xl)Causee (x2)Go HITPAEL· (B7) (xl)Po

(69)c

Bl (+tr) s^ l+v B3 (+tr) B5 (c) l-v |-v B4 (pp) B6

-v (1st B2 (pp)

The predicate-schemes belonging to this diagram PAAL· (Bl) (xl)Ag (x2)Go NIFAL (B2) (xl)Proc PIEL (B3) (xl)Ag (x2)Go PUAL (B4) (xl)Proc HIFIL (B5) (xO)causer (xl)Causee (x2)Go HUFAL (B6) (xl)Proc HITPAEL (B7) (xl)Po

rv (2nd arg) B7 ( r e f )

are:

139

(69)d

Bl (-tr)

^ +v //

B2 (-tr)

\\ +v\ ^\^v (2nd arg)

B3 (c)

B5 (+tr)

-v B4 (pp)

| -v B6 (pp)

B7 ( r e f )

The predicate-schemes belonging to this diagram are: PAAL (Bl) (xl)Po NIFAL (B2) (xl)Po FIEL (B3) (xl)causer (x2)causee PUAL (B4) (xl)Proc HIFIL (B5) (xl)Ag (x2)Go HUFAL (B6) (xl)Proc HITPAEL (B7) (xl)Po

There is only one root where both the NIFAL and the HITPAEL express reflexivity: n . ' . r . 'shake o f f . (70)

aharey sehu yaca mehamayim hakelev nin'ar (B2) / after that-he got-out o£-the-water the-dog shook-itself hitna'er (B7) behozka. shook-itself strongly 'After getting out of the water the dog shook itself vigorously.'

In this case the two patterns are interchangeable. (iv) HITPAEL· reciprocal The last semantic-syntactic function of the HITPAEL· is the expression of reciprocity. The text frequency of the HITPAEL expressing reciprocity in the corpus compiled for this study was lowest in this function. It occurs in the following CSS's:

(71) sCSS 17.5 32.2

Bl

B2

+tr

pp

B3 +tr

B4 pp

B5

B6

c

pp

B7 rec rec

Examples h.b.k. 'hug 1 r.a.h. 'see'

n= 2 2

Semantically the HITPAEL can be derived from either the PAAL or the PIEL active transitive. For example, from the PAAL·: (72)a b

rina Rina 'Rina rina Rina 'Rina

laha§a (Bl) letamar sod. whispered to-Tamar secret whispered a secret to Tamar' vetamar hitlahasu (B7). and-Tamar whispered-reciprocal and Tamar whispered to each other'

Note that in this construction, unlike its English equivalent, it is perfectly normal not to have a Goal. This supports the analysis of 2nd argument reduction.

140

Consider the following examples of second argument reduction from the PIEL: (73)a

reuven Reuven 'Reuven reuven Reuven 'Reuven

b

hibek (B3) et ahiv behamimut. hugged GM brother-his with-warmth hugged his brother warmly' veahiv hithabku (B7) behamimut. and-brother-his hugged-reciprocal with-warmth and his brother hugged each other warmly'.

The CSS's with reciprocal HITPAEL also occasionally feature a causative HIFIL, but there is no reason to assume that the HITPAEL might be derived from it. These CSS's have the following two general schematic representations:

(74)a

Bl (+tr)

-v (Ist/arg) B2 (pp)

I +v ""—^-v(2nd arg) B5 (c) B7 (rec) -v B6 (pp)

The predicate-schemes belonging to this diagram are: PAAL (Bl) (xl)Ag (x2)Go

NIFAL (B2) (xl)Proc HIFIL (B5) (xO)Causer (xl)Causee (x2)Go HUFAL (B6) (xl)Proc HITPAEL· (B7) (xl)Po

(74)b

B3 (+tr)

^^XT7 (2nd arg) B7 (rec)

-v (Ist/arg) B4 (pp)

The predicate-schemes belonging to this diagram are: FIEL (B3) (xl)Ag (x2)Go PUAL (B4) (xl)Proc HITPAEL· (B7) (xl)Po

5.5.1.4 HITPAEL resulting from qualitative valency change In the cases treated so far the HITPAEL·, whether inchoative, reflexive or reciprocal, and mostly when intransitive, was derived by second argument reduction. In the discussion of tables 10 and 11, however, it has been pointed out that there are also cases when the HITPAEL is not derived by valency reduction, but by a qualitative valency change. The CSS's in question are:

(75) SCSS

Bl

15.2 17.2 24.3 44.2

-tr -tr -tr

B2

B3

-tr -tr/pp pp +tr

B4

pp pp

B5

B6

B7

Examples

-tr -tr -tr -tr

g.u.r. h.l.h.l. r.k.k. h.u.l.

'live 1 'dribble 1 'spit 1 'take place 1

n=

13 5 2 16

141

Consider several examples with such verbs: (76) a b

anahnu garim (Bl) kan me-1950. we live-lp.pl here since 1950 'He live here since 1950.' anahnu mitgorerim (B7) kan me-1950. we live-lp. pi here since 1950 'We live here since 1950.'

The verbs in (a) and (b) have the same basic meaning, but differ in Aktionsart. In the HITPAEL· there is an emphasis on duration over a period of time, whereas the verb in the PAAL pattern simply describes a fact.

(77) a b

basibolim

na'u (Bl)

the-ears-of corn 'The ears of corn haSibolim the-ears-of-corn 'The ears of corn

moved-3p.pl in-the-wind moved in the wind.' hitno'e'u (B7) baruah. moved-3p. pi in-the-wind moved in the wind*.

baruah.

Here too, the two forms of the verb n.u.'. 'move* have the same basic meaning but differ in Aktionsart: in the HITPAEL it expresses iterative or durative, whereas in the PAAL it simply denotes a fact. Similar differences of Aktionsart hold also for most of the other verbs in the CSS's listed in (75). The verbs in the HITPAEL· in these CSS's are derived from the PAAL or PIEL (the latter only in 17.2) on the basis of the criterion of morphological simplicity. Since both the input and the output forms have valency 1 and differ in Aktionsart, I suggest seeing these derivational processes as a change of qualitative valency. This use of the term 'qualitative valency change' is used in a broader sense than in Dik 1985, where qualitative valency change specifically involves a change in the semantic function or selection restrictions of at least one of the arguments. In (76) -(77) there is no such change.

5.5.2 Valency increase Bioyania derivation through valency change need not necessarily involve valency reduction but may involve also valency increase. This occurs between two pairs of binyaaia: (i) PAAL· — > HIFIL (ii) PAAL· — > PIEL· 5.5.2.1 PAAL — > HIFIL The prototypical case of valency increase is the formation of the morphological causative. Valency may increase from 1 to 2, as in (78)

cahak 'laughed' yaSav 'sat' ba'ar 'burned'

or from 2 to 3 as in

-

hichik 'made laugh' hoSiv 'made sit hiv'ir 'made burn

142

(79)

axal 'ate'

- heexil 'made eat,

fed'

Sam 'a 'heard' - hiSmi'a 'made hear 1 lavaS 'wore' - hilbiS 'made wear, dressed'

This process is decribed in chapter 4 as predicate formation. Let us look here at those cases where the prototypical valency increase relation between the PAAL and the HIFIL does not apply, i.e. at the CSS 's containing a HIFIL pattern but no PAAL. (Since this valency increase process is optional, analysis of the cases where12 there is a PAAL but no HIFIL will not yield any relevant information.) (80)

CSS 's with B5 without Bl

sCSS 10.1 10.2 16.1 16.2

Bl -

B2

B3

B4

B5

c +tr c c

+tr -tr

+tr

28

21.1 21.2 27.1 27.2

-

34 1 34.2 41

-tr i

B6 pp pp

Examples g.r.l. 'draw lots' p.g.z. 'bomb* h.p.a. 'protect 1 d.l.g. 'jump 1

pp

d.b.r. 'destroy insects'

c c

pp pp

+tr/i pp

-tr

c

pp

-tr

m.r.c. h.l.d. r.s.m. l.a.m.

'zeal' 'rust' 'impress* 'nationalize'

n= 71

13 2 7

19 1 6 1

-tr -tr

'.n.g. 'enjoy 1 1 m.c.'. 'finish s.l.h. 'shed*

18 1 6

PP PP

i/-tr

'.c.m. 'huge' x.t.r. 'crown 1

15 5

PP PP PP

-tr ref i

y.c.v. 'stable, put' h.b.a. 'hide 1 s.r.S. 'root'

15 2 2

-tr

c/+tr pp c +tr pp +tr +tr c

PP

49 52

-tr

+tr +tr

PP PP

c c

57. 57. 57.3

•tr -tr -tr

+tr +tr +tr

pp c pp c pp c

-

B7

Possible sources for the derivation of the HIFIL in (80) are 1. the PIEL pattern as in CSS's number 16, 28, 34, 49; 2. the NIFAL pattern as in CSS's number 41, 42, 52, 58; 3. nominal forms as in CSS number 10.

The HITPAEL, which results from 2nd argument reduction is an unlikely source for the HIFIL· because that would require the reintroduction of the reduced argument first. Clearly there is no good reason for assuming such a process. 1. FIEL as basic pattern Whenever both the PIEL and the NIFAL occur in the same CSS, the basic pattern will be the PIEL, it being morphologically the simplest. The derivations in the CSS's listed above are a. PIEL (+tr) > HIFIL· (+tr/c) PIEL (+tr) PIEL· (-tr)

> >

HIFIL· (+tr/c) HIFIL· (c)

Both PIEL and HIFIL are causative with the roots '.n.g 'enjoy* and a.d.h. 'evaporate*. In isolation both 'ineg and he'enig mean 'cause pleasure' and both ida and heeda mean 'cause to evaporate*.

143

The second type with both PIEL (+tr) and HIFIL (+tr) applies for example to the verb m . c . ' . 'finish, exhaust', or b.r.S. 'brush'. Here too, there is no semantic difference between bereS and hivriS, both meaning 'brush 1 (+tr). The FIEL is chosen as basic because it is morphologically simpler. In the third group the FIEL is intransitive. (This is the only group of verbs where the FIEL pattern is not transitive or causative.) This group consists of verbs like dileg 'jumped' or rooea 'sang and danced'. The derivational process is simply valency increase. In another group of CSS's there are in principle two possible sources for the HIFIL·: the FIEL (+tr) or (c) or the NIFAL (-tr) or (i). Some of1 the roots concerned are: s.l.h. 'send', p.h.m. 'carbon', b.b.a. 'hide , y.c.v. 'put, stabilize' and S.r.S. 'root'. From the semantic point of view, the HIFIL (c) could be derived in all these cases either from the NIFAL or from the FIEL. Since there is no semantic objection to deriving the NIFAL· from the FIEL by 2nd argument reduction, the FIEL is the basic pattern for the CSS, (on grounds of morphological simplicity) and consequently also for the HIFIL. The following diagram represents this:

(81)

B3 (c/+tr) -v(2nd arg.)/ -v(2nd arg.)X^y(lst arg.) B2 (-tr/i)

B7 (-tr/i)

B4 (pp)

B5 (c) -v B6 (pp)

The predicate-schemes belonging to this diagram are: FIEL (B3) (xl)Causer (x2)causee or: (xl)Ag (x2)6o NIFAL (B2) (xl)Po PUAL (B4) (xl)Proc HIFIL· (B5) (xO)Causer (xl)Causee [(x2)Go] HUFAL (B6) (xl)Proc HITPAEL (B7) (xl)Po

2. The NIFAL as basic pattern The fourth type of derivational relation is where except for the HIFIL, there is no transitive pattern at all in the CSS (i.e. no PAAL or PIEL, number 21). Since the NIFAL and HIFIL are morphologically equally complex, the criteria of semantic simplicity and of consistency with the rest of the system remain to be considered. We get the following three cases: (i) HIFIL· with intransitive NIFAL· The verbs in question are: (82)

nimar (B2 -tr) 'got changed' nimrac (B2 -tr) 'hurried' nisat (B2 i) 'got moved'

hemir (B5 +tr) 'changed 1 himric (B5 c) 'made hurry' hesit (B5 +tr) 'moved, 'instigated'

In fact such cases have already been discussed in connection with the NIFAL. When the NIFAL· is linked with the HIFIL it almost invariably means 'result of the action designated by the HIFIL 1 . For this, and other reasons (see discussion on NIFAL· in section 4.5.1.1.), it is preferable to derive it from the HIFIL via argument reduction, rather than the HIFIL· from the NIFAL· via valency increase. Thus, in the cases when the HIFIL is (+tr), it is the basic pattern of the CSS.

144

(ii) HIFIL with inchoative NIFAL There is only one CSS with the biayaaia functions HIFIL (c) and NIFAL (inchoative); for example with the roots m.r.c. 'zeal' or h.l.d. 'rust. Since the HIFIL is transitive, it is the basic pattern of the root, the NIFAL being derived from it by means of detransitivization. This choice is based on the criterion of consistency with the rest of the system. (In the other cases the NIFAL· was also derived from the HIFIL, not the other way around.) (iii) Denominal HIFIL Finally,there are two CSS's without any possible source pattern: numbers 10 and 27. In CSS 27 the roots r.s.m. 'impress' and l.a.m. nationalize* occur in the HIFIL, HUFAL and HITPAEL patterns, and in CSS number 10 g.r.l. 'draw lots' and p.g.z. 'bomb 1 are in the HIFIL and HUFAL patterns. The HIFIL· is inchoative, causative or transitive, the HUFAL is pseudo-passive and the HITPAEL· is intransitive. The problem here is that all the functions are derived, but there is no basic pattern, at least not one capable of being input for any of the valency change rules. In these cases the verbs are derived directly from the nouns of the same root, for example (83)

pagaz 'bomb' - hifgiz 'bombed' hagrala 'lottery' - higril 'drew a lottery' roSem 'impression'- hiraim 'impressed'

The derivational rule here involves a change of the category of the input predicate, i.e. it is not a rule involving valency change. The other deviation from the prototypical valency increase relation is when both the PAAL and the HIFIL· have valency 2. This we find in the following CSS's:

(84) sCSS 13.2 26.2 46.2

Bl B2 +tr +tr +tr pp

B3

B4

+tr +tr

B5 +tr +tr +tr

B6 pp

B7

Examples n= m.r.h. 'disobey' 3 b.d.a. 'invent' 2 k.r.n. 'radiate, shine' 1

Note that although the HIFIL· here has valency 2, like the PAAL, it is derived from the PAAL which is morphologically simpler. The number of CSS's where both PAAL and HIFIL are (+tr) is rather small. The verbs in the two patterns are often used interchangeably, as illustrated below: (85)a b

hanura karna (Bl) beor adorn, the-lamp shone 3p.s.f. in-light red 'The lamp shone in a red light', hanura hikrina (B5) or adorn. the-lamp radiated 3p.s.f. light red 'The lamp shone in a red light'.

5.5.2.2 PAAL

> PIEL

The other valency increase relation is (-tr) PAAL· > (+tr) PIEL·. In the traditional grammars the PIEL pattern is presented as an addition of the feature [+intensive] to the PAAL·. This is true only for a limited group of roots. The classic example is

145

savar (Bl) 'broke 1

(86)

-

Siber B3 (+tr/c) is in line with the prototypical transitivzation process dealt with in chapter 4. In the CSS's where this occurs, there is no limitation on the semantic-syntactic function of the other binyanim. Therefore, we need only look at the summary of the CSS's where such valency increase occurs:

(89)

Bl B2 -tr -tr/ pp

B3 +tr/ c

B4 pp

B5 c/ +tr/ i

B6 pp

B7 i/ -tr/ (ref, rec)

In most cases the valency increase is from intransitive PAAL to transitive PIEL·; in some cases it is from an intransitive PAAL to a causative PIEL· (depending on the meaning of the verb). In all cases it entailes an increase from valency 1 to valency 2. The verbs which undergo this valency increase can be divided into the following two groups: 1. The only difference between the verbs in PAAL and PIEL concerns the quantitative valency. 2. There is also a difference in the qualitative valency of the verbs. 1. Quantitative valency difference In these cases the PIEL is the transitive of the root and the PAAL· is the intransitive, without any further semantic differences. Some examples of such roots are: (90)

patam (Bl) 'force feed' kava (Bl) 'got extinguished' yag'a (Bl) 'tired' (-tr) nasar (Bl) 'fall off

-

pitem (B3) 'force feed' (+tr) kiba (B3) 'extinguished' yige'a (B3) 'tired' (+tr) niser (B3) 'shed'

Here the criterion for the basic binyan is that the PAAL· is unmarked. The PIEL· is derived from it by increase of valency. This is supported by the existence of a causative HIFIL in these CSS's. The HIFIL as causative occurs usually with verbs whose active form is PAAL·

146

(usually intransitive). Except for CSS's number 28, 34.1 and 40, verbs whose active form is in the FIEL do not have a morphological causative. Therefore, it can be assumed that the basic binyan in these CSS's is the PAAL and not the PIEL. Several other examples of such roots are: (91)

dalal (Bl) 'got less crowded'

duel (B3) 'to thin out'

za'af (Bl) 'was angry" yag'a (Bl) 'tired, laboured'

zi'ef (B3) 'angered' yige'a (B3) 'tired' ( + t r ) '

hidlil (B5) 'caused to get less crowded' hiz'if (B5) 'made angry' hogi'a (B5) 'made tired'

2. Quantitative and qualitative valency difference There are also pairs of PAAL - PIEL where there is a semantic difference between the two forms, as in nataf (Bl) 'dripped 1 nitef (B3) sar (Bl) 'move aside' siyer (B3) dag (Bl) 'fished' diyeg (B3) hasar (Bl) 'was missing* - hiser (B3)

(92)

'dripped' (+iterative) 'patrolled' 'fished', (-i-iter.) 'deducted'

In such cases too the PIEL is derived, either because it is in some way semantically more complex or because it is further removed from the basic meaning of the root than the PAAL·. Since all other possible CSS's with a transitive or inchoative HIFIL have a causative FIEL, and this is itself derived from the PAAL, all the cases of valency increase of PAAL· —> PIEL have been accounted for. The other group that forms an exception on the valency increase are those CSS's in which both the PAAL and the FIEL are transitive, i.e. have valency 2. These are listed below: (93) CSS's with both PAAL and FIEL (-Kr) sCSS Bl B2 14.1 +tr

B3 B4 B5 +tr(inten)

20.1 22.2 22.6

+tr pp +tr +tr

+tr(inten) +tr pp +tr +tr

30.1 36.1 37.2

+tr pp +tr pp +tr

+tr +tr +tr

39.2

+tr

+tr

42. 1 +tr PP 42. 3 +tr PP (int) 42. 4 +tr PP 44. 1 +tr PP 45. 1 +tr PP 46. 1 +tr 47. 1 +tr PP 50. 2 +tr

+tr +tr

B6

PP

B7

-tr -tr

+tr/c PP c +tr +tr +tr c +tr PP c +tr +tr PP c

-tr ref PP PP PP

n= 23

c.r.m. •disturb' r.h.t. •furnish' g.h.k. 'burp'

21 15 2

y.z.m. 'take initiative'46 S.f.x. 'pour' 14 d.v.v. 'make talk' 2 h.n.h. 'to park'

PP PP PP

Examples S.b.r. 'break'

-tr -tr

l.m.d. 'learn' g.r.d. 'scratch'

36 1

p.t.h. '.z.v. d.k.r. n.g.h. p.r.s. t.l.a.

8 21 19 3 16 43

Open' 'leave' 'jab' 'jostle' 'slice 1 'patch'

147

Cont. table (93): sCSS 52.1 52.2 53.1 54.1 55.1 59.1 60.1 60.3 60.6 60.8

Bl +tr +tr +tr +tr +tr +tr +tr +tr +tr +tr

B2 pp i pp pp pp pp pp pp ref

B3 +tr +tr +tr +tr +tr +tr +tr +tr +tr +tr

B4 pp pp pp pp pp pp pp pp

B5 c c c c c c c c c

B6 pp pp pp pp pp pp pp pp pp

B7 -tr i -tr -tr -tr -tr i ref ref

Examples s.t.m. ' f i l l 1 k . m . r . 'round' r.c.h. 'want' S.m.n. ' f a t , oil' b.d.k. 'check' y.s.d. 'found 1 m.c.a. 'find 1 m.n.h. 'appoint 1 s.r.k. 'comb 1 n . ' . r . 'shake off

n= 48 15 16 28 25 6 64 22 5 1

The number of CSS's in which both PAAL· and PIEL are transitive is quite large. In some cases there is a semantic difference, the PIEL being intensive (see example (86) above) or durative, as with (94)

halax (Bl) 'walked' -

hilex (B3) 'walked (durative) 1

In other cases the two verbs have different meanings, as with (95)

lamad (Bl) 'learned 1 - limed (B3) 'taught'

Mostly, however, the PAAL and the PIEL are synonymous. For example (96)

gasar yasad sarak na'ar 'akac pacah

(Bl) (Bl) (Bl) (Bl) (Bl) (Bl)

-

giser (B3) 'bridged, connected 1 yised (B3) 'founded 1 serek (B3) 'combed' ni'er (B3) 'shook off 'ikec (B3) 'stung 1 piceah (B3) 'broke open 1

With some of these synonymous verbs the PIEL is more frequently used than the PAAL, as in (97)

garad (Bl) - gered (B3) 'scratched' sakal (Bl) - sikel (B3) 'stoned' pacah (Bl) - piceah (B3) 'broke'

The PAAL is assumed to be the basic pattern for the whole group on the basis of the criterion of consistency and the fact that in part of the cases the PIEL is semantically marked. In some instances the PIEL is derived from it by a change in the qualitative valency. Otherwise, no change of valency occurs. Let us now return to the third sub-hypothesis presented at the beginning of this chapter, namely that the derivational relations between the binyanim in the CSS 1 s can be described in terms of valency change, similar to those used in chapter 4 to describe the productive processes. Because we are dealing here in principle with the small, "irregular" group of verbs, this hypothesis, like the two other hypotheses, should not be expected to be borne out by each and every case. Nevertheless, the changes in the quantitative valency in the CSS's do hold good for most cases. In addition, the pairs which do not undergo quantitative Valency change, do often undergo qualitative valency change. The hypothesis is weakened by the existence of pairs which seem synonymous, though often one of the pair is more frequently used than the other. There may be a

148

pragmatic difference in the use of the binyanim in such pairs. That, however, is a complex topic, demanding a study in its own right. 5.6 Summary In this chapter I have attempted to suggest a level of organization, lower than that of the overall verbal system, where I expected to find regularity in the functions of the binyanim and in the derivational relations between them. This was partly confirmed. No considerable degree of consistency could be found in the syntactic-semantic functions of the seven binyanim. Nevertheless, several interesting insights were found in the derivational relationships between them. These will be discussed below, first, regarding the Modern Hebrew verbal system, and next with regard to the predicate formation mechanism in Functional Grammar. 5.6.1 Regarding the Modern Hebrew verbal system The major hypothesis regarding the CSS's is divided into three connected sub-hypotheses: the consistency of the semantic-syntactic functions of the binyanim in the CSS's, the consistency of their valency in the CSS's and the possibility to describe the derivational relationships between the binyanim in terms of valency change. The first hypothesis holds only for a small part of the CSS's, but the second and the third hold for the majority. In fact the first two sub-hypotheses should be seen together, in view of the argument stated in chapter 4 that valency is more suitable for describing the HH verbal system than transitivity. That is, the first hypothesis is based to a large extent on transitivity and not on valency as the second one. It turns out, however, that the consistency in valency within a CSS is higher than in semantic function. Thus, by and large the three hypotheses seem quite valid. Furthermore, a closer look at the derivational relations in HH viewed in the context of CSS's reveals new information. Some of the derivational relations coincide with the ones described in chapter 4 for the productive processes, others differ. The ones similar to the productive, prototypical processes are: -v 1. Bl (+tr) > B2 (-tr/pp/ref) (Eg. in CSS's 19, 24, 29, 32, 36, 43 etc) This valency reduction relation applies regardless of the function of the other binyanim in the CSS. The CSS's in which there is Bl (-tr) (in that case the B2 is active intransitive) are exceptional. +v 2. B3 (+tr/c) > B4 (pp) (Eg. in CSS's 8, 17, 30, 43 etc) This valency reduction relation holds for all CSS's featuring both B3 and B4. -v 3. B5 (c/pp) > B6 (pp) (Eg. in CSS's 10, 29, 35, 39, 45 etc.) This valency reduction relation holds for all CSS's featuring both B5 and B6. 4. General 2nd argument reduction: -v Bl (+tr)/ B3(+tr) /B5(+tr) > B7 -v a. Bl(+tr) /B3(+tr/c) —> B7 (-tr)

149

-v b. Bl (+tr) /B3(+tr) /B5(+tr) ---- > B7(i/resultative) -v -v c. BK+tr) /B3(+tr/c) ---- > B7 (ref) ; B5 --/— > B 7 ( r e f ) -v -v d. BK+tr) /B3(+tr) ---- > B7 (rec) ; B3 ( c ) / B 5 ( + t r / c ) --/-- > B7 (rec)

5. Valency increase: causativization +v --- > B5(c) The CSS 's proved useful when looking at the derivational relations which deviate from the prototypical processes. These deviating processes are the following: 1. When there is no Bl in a CSS, B2 can be derived also from B3 (+tr/c) or B5 (+tr/c) ; B2 then has a resultative meaning. Thus we get: -v B3 (c/+tr) — > B2 (res) -v B5 (c/+tr) — > B2 (res) The NIFAL (B2) can also be derived not only from its prototypical source pattern, the PAAL, but also from the PIEL, in particular when the PAAL is (-tr) , according to the following process: -v B3 (c/nr) — > B2 (res) Similarly, the HUFAL (B6) can be derived not only from the HIFIL, its prototypical source, but also from the PIEL, according to the following process: -v B3 (+tr/c) — > B6 (pp) Other cases of valency reduction which do not follow the prototypical relation are the derivation of the PUAL from the HIFIL· (instead of the PIEL), and of the NIFAL from the HIFIL instead of the PAAL. The latter is a special process, because the NIFAL derived from PAAL is (pp) , whereas when it is derived from the HIFIL it is resultative. These derivational processes are: -v B5(+tr) ---- > B4(pp) -v B5(c/+tr) — > B2(res) (in particular with Bl (-tr) ) Hence, a detransitivized binyan can also be derived from its nonprototypical source, such as NIFAL· from HIFIL·. Other deviations occurred when both binyanim of a pair had the same quantitative valency, as for example when both PAAL· and NIFAL· were intransitive. In some of these cases one of the pair can be derived from a different binyan than the prototypical source (eg. NIFAL· from HIFIL) . Alternatively, there may be a difference in the qualitative valency or in markedness. Such is the derivational process of deriving the HIFIL· (c) from the PAAL (+tr) . Here often there is no change in the quantitative valency (i.e. when the HIFIL is bivalent), but in the qualitative valency: qual. change Bl(+tr) ------------- > B5(c) The diagrams below summarize the possible derivational relationships (they do not represent actual CSS 's):

150 Bl(+tr)

(98)a -v (ls$/irg)

-v (2nd arg)~"\ +v

B2(pp/ref)

B7(-tr/i/ref/rec)

B5(c) -v(lst arg) B6 (pp)

Bl(+tr)

-v(lstxarg)

-v(2nd arg)

B2(pp)

B5(c)

B3(+tr)

B7(ref/rec/i/-tr)

-vilst arg)

-v(lst arg)

B6(pp)

B4(pp)

Bl(-tr) y-v(2nd arg] B2(-tr/i)

B3(c)

B7(-tr)

+v

B5(+tr) -v(lst arg)

-vdst arg)

B6(pp)

B4(pp) B3(+tr)

^ I \ -vdst -v(lst argLx^vdstyargJX

^

B2(i/res)

B2(i/res)

1 \

B4(pp)

B6(pp)

arg) B7(i/res/rec/ref)

B6(pp)

The underlying predicate-schemes in all these cases are the standard ones: PAAL (Bl, +tr) (xl)Ag {x2)Go or PAAL (Bl, -tr) (xl)Po VlfKL (B2, pp) (xl)Proc or (B2, -tr/i/ref/rec) (xl)Po

151

FIEL (B3, +tr) (xl)Ag (x2)Go or FIEL (B3, c) (xl)Causer (x2)Causee PUAL (B4, pp) (xl)Proc HIFIL (B5, c) (xO)Causer (xl)Causee [(x2)Go] or HIFIL (B5, +tr) (xl)Ag (x2)Go HUFAL (B6, pp) (xl)Proc HITPAEL (B7, -tr, i, ref, rec) (xl)Po Note that (98d) and (98e) represent only potential valency reduction; both B3 and B5 can undergo 1st argument reduction leading to either B4 and B6, their pseudo-passive equivalent, or to B2. When a B3 or B5 verb has two reduced equivalents, the B2 will usually be resultative, and B4 and B6 pseudo-passive. The B3 and B5 can also undergo 2nd argument reduction, leading to B7 inchoative or resultative. Never do all four reduced binyanim occur together, nor do I intend to suggest that they do. Finally, there are also binyanim which are derived directly from a nominal predicate: (99)f

C.C.C. N (xl)0

>

C.C.C. V-B7(-tr/i)

In total there are 60 CSS's and 110 sub-CSS's (including the CSS's with one sub-CSS only). In general it can be concluded that the level of CSS is meaningful for the Modern Hebrew verbal system. 1. It gives a clearer picture of the derivational relationships between the binyanim than the one obtained by looking at the whole population of roots. Furthermore, it accounts not only for a minority of roots occurring in all 7 binyanim. 2. The context provided by CSS's helps determine the direction of a derivation in cases when theoretically there are two alternatives. 3. The CSS context helps to identify the basic binyan when in principle there is more than one candidate. 4. If valency changes are regarded as predicate formation rules they apply simultaneously to the syntactic and semantic levels, (as both are represented in the predicate-frame). This way an apparent syntactic redundancy can be explained: sometimes two detransitivised binyanim of the 1st argument reduction type, B4 and B2 or B6 and B2 occur as derived from B3 or B5. There is, however, a semantic difference between them: B4 and B6 are pseudo-passive, whereas B2 in such CSS's is resultative. This chapter leads to the following conclusions with regard to the NH verbal system: a. Although there are many regularities which hold for small groups of roots, the final picture is certainly less chaotic than often presented in the literature. b. The notion of binyanim has proved its value as the organizing principle of the MH verbal system. I wish to point out that these CSS*s and the regularities for the derivations of the binyanim do not stand on their own, but are in fact part of the lexicon. Therefore, it seems that the picture of the MH verbal system which finally emerges from the lexicon (chapters 3, 5) and the productive predicate formation rules (chapter 4) is coherent and unified. Thus the strength of this analysis is its unity gained by using the same principle of valency change as the basis for all the relations. A possible objection could be that it is arbitrary to distinguish between the same types of processes as being sometimes productive and sometimes non-productive. In practice, however, sometimes, the

152

non-productive processes concern other relations between other binyanim than the productive processes do. (e.g. the derivation of the NIFAL from the HIFIL, or of the FIEL from the PAAL.) This point will also be discussed in chapter 6.

5.6.2 Regarding Functional Grammar The general contribution of the results presented in this chapter to the theory of Functional Grammar will be discussed in the next chapter, together with the other results. Here I wish to point out a problem raised by the preceding discussion: is it useful in the context of the Modern Hebrew verbal system to distinguish at all between the predicate formation rules described in chapter 4 and the regularities described in this chapter? Chapters 4 and 5 show that it is possible to describe the derivational relationships between the binyanim as processes of valency change which were mostly quantitative and in some cases qualitative. Since we are talking about the same processes, it seems arbitrary to divide them into productive and non-productive. One could also say that the productivity is contained in the size of the CSS, there being two big CSS's, (FIEL- PUAL and FIEL- PUAL-HITPAEL), which are also open and productive, while the other CSS's are non-productive. One answer is that this distinction is based on the methodology of FG, which is the theoretical framework of this study. Another possible answer would be that some of the derivational relations applying to these groups are definitely seen by the speakers of contemporary Hebrew as productive, such as the causative formation with the HIFIL pattern. The other issue raised in this chapter is the definition of productivity in Functional Grammar. A rule needs to meet two demands to be considered productive in Functional Grammar: it should be applicable to a large and open group of items; and the speakers of the language should know and use it as such. The processes as described in chapter 4 meet the second demand; only two of them meet the first demand. The same processes which do have psychological reality, and on that basis can be seen as productive, apply also to small groups, which have to be learned as such, and are therefore non-productive. Hence, at least for the verbal system of Modern Hebrew it would be much simpler to say that the system consists of 2452 roots organized in CSS's, and that the binyanim in which each root occurs are listed in the CSS's, as are the processes by which they are related to each other. Consistent application of this principle would result in the elimination of the distinction between lexicon and predicate formation rules, at least in so far as the verbal system is concerned. Note that this line of argument is based wholly on the quantitative criterion (i.e. the group being large and open) of productivity. One could claim that there are intuitive, and also empirical reasons for assigning greater importance to the psychological criterion, and for retaining the distinction between rules and regularities.

Chapter 6

Conclusions

6.1 Result· This study had a dual aim: on the one hand it set out, to describe the verbal system of Modern Hebrew using the theory of Functional Grammar. It was assumed that F6 would provide a new perspective, offering new insights into the Modern Hebrew verbal system. On the other hand, the Modern Hebrew verbal system was projected as a test case for the theory of Functional Grammar: more specifically, for part of its fund: the lexicon and the predicate formation mechanism. With regard to the first aim, two specific hypotheses were formulated: a. It is possible to divide the whole verbal system into configurational sub-systems, and the biayaaim, whose functions and derivational relationships are highly irregular on the level of the overal system, are much more regular on the level of the CSS's. b. The derivational relations between the bioyania, both the productive and the non-productive, can be described in terms of changes in the valency of the binyaaim. Mostly these changes are quantitative, and in some cases they are qualitative changes. The second aim was to test whether Functional Grammar could account for the roots and for the whole phenomenon of roots and binyania, since to date there is no FG description of languages containing such a mechanism. 6.1.1

The Hebrew verbal system

The hypothesis regarding the regularity of the functions of the binyanim and their derivational relations on the level of CSS's was further specified in three sub-hypotheses: (i) the semantic-syntactic functions of the binyaaia within a CSS are consistent. This hypothesis turned out to be valid only for a small number of CSS's. In most CSS's the binyaoim which are potentially multi-functional (the PAAL, NIFAL, FIEL, HIFIL and HITPAEL) have more than one function. Two points are noteworthy in this respect: first, that the more bioyaaia the CSS contains, the more functions the multi-functional biayanim have. Exceptions are the very large and the big medium-sized CSS's, where the multi-functional binyania show less diversity of functions. But the difference not significant enough to warrant the conclusion that the diversity in functions is limited to the smaller CSS's. Second, although the multi-functional binyaaim were not consistent, they practically never fulfilled all their possible functions in the same CSS. (ii) the quantitative valency value of the multi-functional binyaaim is consistent in the same CSS. This sub-hypothesis turned out to be correct for a much larger number of CSS's than the first one. In fact it applied to the majority of the CSS's. The importance of this result is that it adds considerable weight to the claim that the notion of valency is essential to the description of the Hebrew verbal system. Whereas the semantic-syntactic functions, like passive, causative, reflexive, reciprocal, inchoative etc.

154

emphasize the meanings associated with a verb in a certain binyan, the notion of quantitative valency emphasize the combinatorial possibilities of the binyanim. Thus, the meanings associated with the specific binyanim in CSS's may be a good deal less consistent than their combinatorial possibilities. The latter also implies a fairly high degree of consistency in the types of constructions in which a root belonging to a certain CSS can occur in one of the multi- functional binyanim. Furthermore, the semantic-syntactic function depends on every separate root + binyan combination, whereas the valency does not. Hence, valency is a better way of abstracting the features of the binyanim. (iii) the derivational relations between the binyanim both in the CSS's and in the productive part of predicate formation rules can be described in terms of valency changes. This hypothesis turned out to be valid for most cases. In some of the cases where the derivational relation cannot be described by means of quantitative valency change of the binyanim, it can be described by qualitative valency changes. Some cases remained where neither of the valency change processes offered an adequate description. These cases possibly involve pragmatic factors of usage, which, however, were not checked in this study. In addition, the fact that practically all the derivational processes, both the productive and nonproductive, can be described by means of valency change gives this description a considerable unity. Classifying all the roots in CSS's has the advantage of presenting a clear picture of the distribution and functions of the binyanim in the overall system (and not just the major tendencies, as most studies do). Thus for example we obtained the regularity that when a root occurs in the NIFAL pattern and does not occur in the PAAL (+tr) but does occur in the HIFIL (c), it expresses resultative and not pseudo-passive. Another regularity found was that a root in the HITPAEL derived from the PAAL or PIEL pattern expresses intransitivity, reflexivity or reciprocity; but when derived from the HIFIL (c) it expresses resultative (or inchoative). The CSS's may have the additional advantage of providing a partial explanation for gaps in the system (i.e. for the fact that some roots do not occur in certain binyanim although there is no semantic reason for this). The distribution of roots in the CSS's shows regularity in occurrence, but does not give any information on causal relationships. Nevertheless, the principle on which the distribution is based can provide some explanation for gaps in the system. It is possible that some roots owing to their meanings express a specific function in a certain pattern. If this pattern, however, is derived, and the input is missing, this explains why the roots in question do not occur in these patterns. Another type of gap in the system are CSS's which are theoretically possible, but which do not occur in practice. This, as shown in chapter 5 section 5.3, can indeed be explained by the principle of qualitative valency change. For example, no roots occur in the various combinatorial possibilities of NIFAL, PUAL, HUFAL and HITPAEL when these are the only binyanim in a CSS, because all of them are reduced binyanim which require a bivalent input pattern in the CSS. That is, some roots which could be meaningful in a certain pattern, do not occur in that pattern because it is derived, but lacks the potential input. The other contributions to the description of the verbal system of Modern Hebrew resulting from this study are the predicate-schemes and the lexicon. The predicate-schemes are predicate frame- like entities, but feature a binyan instead of a concrete predicate. The predicate-scheme reflects the number of arguments the binyan can take, as well as their semantic functions. Selection restrictions on the arguments can be added

155

only when discussing concrete verbs (i.e. binyan + root). The validity of the predicate-schemes in the description also supports the validity of the notion of binyania as the underlying principle of the MH verbal system. The second contribution is the proposal of a lexicon, with entries consisting of CSS's followed by a list of the roots that belong to the CSS. Each root is followed by the nominal forms that can be derived from it. The resulting lexicon will be much more highly organized than other lexicons of MH (eg. Herman (1975a), Schwarzwald (1975)) while having the additional advantage of being organized on the same principles as the rest of the description. 6.1.2 Functional Grammar The present study also contributes to the theoretical framework of FG: it enlarges the descriptive mechanism at certain points, and it identifies some problems. The descriptive contributions are the notion of the predicate-scheme and the lexicon. The predicate-schemes are in fact an extension of the predicateframes. They provide substantial proof for the validity of the concept of pedicate-frames. As the predicate-frames are unsuitable for describing the binyanim, they also enable the extension of the concept to the binyanim system. This allows a compact description of the binyanim in structures that reflect not only their semantic-syntactic functions but also the number of arguments and their semantic functions. The predicateschemes are also applicable to descriptions of Arabic, which has the sane structure of roots and patterns. The lexicon as described in chapter 3 is a more original contribution to FG. In the first place, it is organized on a certain principle, whereas the lexicons suggested in FG so far consist of a list of all the basic predicates of a language. In the second place, the lexicon is structured according to the same principle as the derivational processes: a basic binyan plus the other binyania derived from it, i.e. the CSS's. This gives additional support to the notion of predicate schemes, and the analysis of the derivational processes in terms of valency changes. Another issue raised in this research is the notion of 'productivity' when applied to the binyanim system. In the case of some binyanim there is a clash between the two criteria for productivity: applicability to a large and open group (quantitative criterion) and psychological reality. Only three CSS's could be conclusively said to be productive: PIEL, PIEL-PUAL and PIEL-PUAL-HITPAEL. (CSS number 60 with all 7 binyania is a big one, but the variety in the function of the binyania in it is too big, and no group contains more than 100 roots- which is the limit for a big and productive CSS.) These CSS's meet both the quantitative and the psychological criteria, and are also predictive. According to the quantitative criterion, only the functions of the binyanim in these CSS's and the derivational processes of (+tr/c) — > ( p p ) and (+tr) — > ( - t r ) are productive. But the rule of HIFIL causative formation, for example, is not productive, because it applies to a closed group of roots that have at least in part to be learned. (Because there are many cases where a certain root does occur in the PAAL but does not form its causative with the HIFIL.) At the same time, the use of HIFIL for expressing causativity does have a definite psychological reality, as shown by empirical research (for example Herman (1980) and Bolozky (1983)). There are two possible solutions to this problem:

156

(i)

the cases where the two criteria clash are indeed unclear with regard to transitivity and are either edging towards transitivity, or declining into lexicalization; (ii) the preference of one of the criteria to the other, in the sense of considering one of them more important and taking a decision on the basis of this one criterion only, even if the other one is not fulfilled. Contemporary Israeli Hebrew is a very dynamic language, which is still (almost four generations after it became spoken again) undergoing changes which reach further than the normal lexical additions and deletions. Bolozky (1978, 1983) shows which biayaaia are most widely used for the formation of new verbs. This may mean that in the long run some of the binyania will become obsolete like the NITPAEL. It is conceivable that in the not too far future the binayaim system will be simplified, along the lines of transitivity emerging from the larger CSS's. This, however, does not help to decide which processes are productive in MH at the present time. I have, therefore, opted for possibility (ii) putting psychological reality before size and openness of the group described. Hence, wherever there was a conflict, involving a process perceived by speakers of Hebrew as productive, it was given the status of a predicate formation rule even when applying to a closed group. The rules involved concretely are +v - PAAL

> HIFIL (c)

-v

- PAAL/PIEL

> HITPAEL (ref) -v - PAAL· /FIEL·/ HIFIL ~> HITPAEL· (rec) : a clear borderline case The other issue raised by this study is passive constructions in FG. Passive constructions are seen by Dik (1978, 1979a, 1981) as a matter of perspective. Thus when a State of Affairs is seen from the point of view of the Ag, Force or Positioner, it is expressed by an active construction. This is coded in the underlying predication by assigning Subj to this argument. When the State of Affairs is described from the point of view of the Go or Processed, this is expressed by a passive construction and is coded in the underlying predication by assigning Subj function to the Go or Processed. There are, however, several FG studies which claim that 'the passive 1 is not a matter of perspective, but of predicate formation: i.e. 1st argument reduction. Such claims have been presented by Vet (1986) for French. Risselada (1987) and Moutaouakil (lecture, September 1986, University of Amsterdam) in studies of for Ancient Greek middle-passive and 'passive' patterns in Arabic also see it as a metter of argument reduction. In Moutaouakil's view too, the constructions usually called 'passive' turned out in fact to be reduced constructions (of the type of 1st argument reduction), but the constructions described by them are distinct from what Dik would call 'passives' and derive by Subj assignment to Go. The constructions in question are seen in this study as 'pseudo-passive', as they differ in certain points from the passives 'par excellence' in English, or Dutch for example. In this study (chapter 4) it was shown that the reasons for reintroducing the Ag-phrase are pragmatic, that is that the so called passive, at least in MH is derived by first argument reduction, and not Subj assignement to Go. A similar analysis for the 'passive' in Arabic has been suggested by Moutaouakil (ibid). Bolkestein (1981) suggests to see Subj assignment as expressing perspective, and demonstrates that on a level higher than a predication Subj can express foreground/background.

157

for example. It seems to me that whereas the derivation of the so called 'passive' binyanim is a question of argument reduction, the use of the 'passive 1 binyanim is also connected to inter-sentential perspective as demonstrated by Bolkestein (ibid); this has been argued regarding the use of the biayanim also in Junger (1985b). How the two approaches, Subj assignemnt for expressing perspective (foreground/background) and argument reduction for expressing the pseudo-passive relate to each other- needs further investigation. 6.2 Future perspectives There are several issues which have only been touched on or even left totally out of consideration by this study, but which definitely deserve more attention. Regarding the description of MR, a study of the representation of the nominal system, the rules underlying it, and its representation in the lexicon would certainly be worthwhile. This would provide, for example, a way of representing the root in a predicate scheme-like construction, but without category. Another issue which would link up with this study is the role pragmatic factors play in the binyaoim system. I briefly suggested that in some cases where there is neither a quantitative nor a qualitative difference of valency between two binyanim, there may be a pragmatic factor which determines their distribution. This may also hold for some of the 'gaps' in the system of binyania: when a certain root does not occur in a certain pattern for no obvious reason. Finally, studies of productivity and passives, issues touched on in this study, would provide interesting contributions.

Notes

Notes to chapter 1 1. In the actual words derived from a root one (or two) of the radicals can change or be omitted. Such changes can be consistent throughout all the derivatives, such as the omission of the middle radical /u/ in ä.u.t. 'sail 1 , or occur only in part of the derivations. In the root k.t.b. 'write* the first radical /k/ changes into its unvoiced equivalent /x/ when it is not in word initial or syllable initial position; the third radical /b/ can also become the unvoiced /v/. In fact this latter change occurs in all the words derived from k.t.b. except in k tuba '(religius) marriage contract'. 2. Naturally the verbs featuring a noun from the same root are not denominative. Gesenius decides for each case anew which verbs qualify on the basis of semantic criteria. For example the verb dag 'fished' is derived from the noun dag ' f i s h ' . 3. HUFAL and HOFAL refer to the same pattern; the choice of the names varies in the various grammars. 4. Note two points: a. not all verbs occurring in the HIFIL also have a HUFAL equivalent; b. I know of no case in MR where the HUFAL is the passive equivalent of the PAAL. 5. There are two noteworthy phonological points about the HITPAEL: a. there is a methathesis in the /hit/ with stems beginning with /s, s, c/ so that we get hictadek 'justified himself and not *hitcadek b. the /t/ of the /hit/ is assimilated when the stem begins with / n , x , t , t , d / , so that we get hi t aber 'purify oneself and not *hittaber; this assimilation is sometimes optional, so that we get both hinabe and bitnabe 'prophesy' 6. Below there is an illustration of how the system works for the benoni, quoted from Oman (1971c): The benoni has the following stems: (a) xac (b) xic (c) xe (d) { xoc, xuc) (e) xvc c i i legend: c = consonant, (thus all the groups (except (d) ) end in a closed syllable; - any combination of consonants and vowels. Thus kam 'got up' is represented by group ( a ) , and boS 'ashamed' is represented by group (e); words consisting of more than one syllable are classified on the basis of their last syllable.

159

Group (£) represents cases where on the basis of the various inflected forms the last consonant is doubled, such as dagg 'fish' vs. yamm 'see'. This difference is based on Oman's other morphological rules. These groups are the stems for both nominal and verbal forms. The other rules are (i) the feminine will be marked by the suffix a when the stem belongs to groups ( b ) , (e) or ( f ) ; otherwise the feminine will be marked by the suffix c: S —> a/llCvC + -, XiC + -, XoC + -, XuC + -, XvC C + -, n i i fCaCeC + -, Xe + -J This rule should be read as follows: Sn is the formative for feminine, /a/ or /t/, and in all the empty places marked by - to the right of the ' / ' comes an a. However, there are exceptions to this rule. 7. Weak verbs are those in which one of the consonants of the root is systematically changed or omitted in the concrete words, either in all its occurrences or in some of them. For example, the /u/ in the root S.u.t. 'sail' occurs only in the future tense, being systematically omitted in the past and benoni paradigms. 8. Note that horiSa 'bequeathed' is seen as a lexicalised causative and not as a direct one. But there are many instances where the HIFIL expresses the causative of PAAL. 9. Similarly, Ariel (ibid, p. 223) formulates a rule according to which the HITPAEL can express the resultative Aktionsart, on the basis of the following system: (i)a b c (ii)a b c d

hame'il mekuvac ( a d j ) . the-coat shrunken 'The coat is shrunken* yosef kivec (B3) et hame'il. Yosef made-shrink GM the-coat 'Yosef made the coat shrink' hame'il hitkavec (B7). the-coat got-shrunken 'The coat got shrunken'. hame'il mekucar (adj). the-coat shortened 'The coat is shortened' yosef kicer (B3) et hame'il. Yosef made-short GM the-coat 'Yosef shortened the coat' hame'il mitkacer (B7). the-coat getting-short 'The coat is getting short', hame'il kacar (adj.) the-coat short 'The coat is short'

In Ariel's example non-verbal forms are also part of the system, because they are part of the logical relation expressed by the rule. In my system the roots are the starting point, and in examples (i) and (ii) they would belong to the following two CSS's: a. k.v.c. 'shrink' B3, B4, B7 b. k.c.r. 'short 1 Bl, B2, B3, B4, B7

160

10. One of the reasons given by Schwarzwald here is that only 2.3% of all the roots of MH occur in all the seven binyanim. However, it is not justified to conclude from this that the binyanim functions are inconsistent in all the rest. If a root occurs in say only 4 patterns, it is still possible (and in fact often happens), that these four patterns do fit into the system. Thus for example the root t.r.m. 'donate' occurs only in the PAAL, NIFAL· and HIFIL, but the regularity PAAL1 active : HIFIL causative does hold good. Or, the root y.c.a. 1'exit which occurs in PAAL, PIEL, PUAL, HIFIL and HUFAL, and h.l.x. 'go which occurs in the PAAL, PIEL, HIFIL and HITPAEL still fit into the system. The existence of a consistent syntactic value of the binyanim is also supported by neologisms and child language, despite the fact that there are also errors in the use of the binyanim. One such fairly common error is the use of the HIFIL· as inchoative instead of the PAAL with the root r.z.h. 'thin', where people say hu hirza 'he grew thin' (r.z.h. + HIFIL·) instead of the normatively correct Λu raza 'he grew thin' using r.z.h. + PAAL. Schwarzwald is right, however, in pointing out that in the case of the multi-functional binyanim there is no way to predict which value the binyan will get with a certain root. Note· to chapter 2 1. Here in fact we come to a problem which as yet has not been entirely solved in F6. Namely, that there are no clear criteria determining when a term should be seen as a satellite or as an argument. One such criterion is that if the whole predication falls within the term's scope, it is a satellite; if only the predicate falls within the term's scope it is an argument. This necessarily leads to ad hoc decisions, which, however, is in itself not a problem, it being perfectly legitimate to have a context-dependent criterion. The problem is that it is not clear whether the basic predicate-frame consists only of the obligatory arguments or includes the optional ones. If the former alternative is chosen, when are the optional arguments introduced into the predicateframe? For example in (i)

John wrote the letter with his new pen

is 'his new pen' an optional Instrument argument or a satellite? According to the 'do so' test, whereby a constituent that can be preceded by 'do so' is a satellite, 'his new pen' is a satellite. ('John wrote a letter and he did so with his new pen.') If the basic predicate-frame includes the optional arguments, the resulting predicate-frames are rather long and contain argument slots that are often left empty. This has methodological disadvantages. 2. In principle it is possible to defend an approach that derives the binyanim completely by expression rules. Such an approach, however, would go against some of the basic assumptions of the FG model, whereby semantic and syntactic aspects of derivational processes are treated by predicate formation rules.

161

Notes to chapter 3 1. I am indebted to Y. Tobin for pointing out this innovatory aspect of the distinction between lexicon and dictionary. 2. She rejects the distinction between verbal and nominal patterns, for the following reasons: A. it is assumed that there are certain phonological restrictions that hold for the root as a morpho-phonemic unit; (Greenberg, 1950). According to Schwarzwald, however, these restrictions apply to the level of what she calls 'morphemes' , and not to the root level. B. the assumed consistency and regularity in the verbal system does not stand up. It applies in fact only to two of the bioyaoia: PUAL and HUFAL· which are always passive; as the total system consists of 7 biayania it is only valid for a very small precentage of all the roots in MH. Here Schwarzwald's argument is based on the fact that one of her experiments demonstrated that people cannot always identify the root. This occurs mostly in two categories of roots: (a) homophonous roots, where in different binyaoim we get in fact a different root. For example (ii)

s.f.r. is in PAAL 'count', in PIEL 'tell (a story)' and in HITPAEL 'get one's hair cut'

(b) the so called 'weak roots', where the first or the second root consonant is a glide; this is always omitted in the actual words, so that speakers never encounter it in the concrete language. The psychological reality of roots, however, has been convincingly demonstrated by Waiden (1982). Admittedly, not for the 'weak roots' separately, but for roots in general. Still, just as Schwarzwald's argument against roots + binyanim is based on the weak roots only, a counter argument can be based on the strong roots only, as weel. The following example illustrates the representation of a root + binyania in Schwarzwald's view: (iii)

/brk + pattern 6 / hivrik (Hifil) 'shone 1 - spirantization of k -transitive ^active 1 shone' • ··

etc

The following examples illustrates a lexical (iv)

/l.b.s. + paal/ 'put clothes on body' + redundancy rule 6 + redundancy rule 7 etc

entry:

(HITPAEL = reflexive, j.j.) (HIFIL = causative, j . j . )

This suggestion raises the problem that there are no criteria for selecting the basic binyaa. In the two examples quoted above this does not present a problem. But what about roots like y.'.c. 'advise' where both ya'ac (Bl) 'advised* and yi'ec (B3) 'advised' are ^transitive and synonymous? Which one is basic?

162

3. Thus the root l.b.S. 'wear, dress' is represented in the lexicon as follows (example quoted from Berman (ibid) ) : (v)

1. basic pattern

|NP [+human] [+agent]

NP ) [-(-clothing item] ϊ+accusative] [-passive] ((ie-)NP )

the between the NP's is the place of the verb in the sentence. (That is, we get a syntagmatic representation.) 2. PAUL : stative 3. HITPAEL : iterative 4. HIFIL : causative The root l.b.S. in the PAUL, HITPAEL· and HIFIL has a regular productive syntactic function, and is therefore derived by a redundancy rule, like (iv) above. Berman suggests listing all the idiosyncracies of the verb in the lexicon. 4. Despite the similarity between the two approaches, there are also some differences in the details: a. Schwarzwald sees only the derivations of what she calls the passive biayanim NIFAL, PUAL and HUFAL as regular, and therefore, as derived by grammatical rules. All the other binyanim and syntactic functions are seen as idiosyncracies, since their occurrence with one specific root or another is unpredicatble. Also the syntactic value which multi-functional binyanim like the HIFIL or HITPAEL· will take with a certain root is unpredictable. b. Schwarzwald does not give the whole syntagmatic environment in which a root in a certain binyan occurs. 5. The oversimplification is that the semantic, syntactic and morphological values cannot be completely separated from each other, nor can the root and pattern be concretely separated in an actual word. But in so far as it is possible to look at each component separately, the binyanim are a factor in establishing the syntactic features of a word, whereas the root or the aiSqalim are not; The root determines only the morphological and the semantic aspects, the miSqal mostly only the morphological. 6. It is impossible to represent the regularities in the miSqalim system too because they cut across the CSS's. That means that the groups of roots which the miSqalim regularities apply to (for example of maCCeCa or CoCCan for Instrument) do not overlap with any CSS. Because this regularity is rather low (I estimate it to be 10-15% of the miSqalim) I have chosen to ignore it here. This is a drawback of the lexicon presented here, but it is not really crucial as this study focuses on the verbal system. 7. r.u.c. is a weak verb and in all its nominal forms the middle radical /u/ is omitted.

163

Notes to chapter 4 1. The majority of examples in this chapter and in chapter 5 come from a corpus specially compiled for this study. It consists of 561 sentences, from the following sources: - radio programmes, where listeners phone to consult an expert - 2 news reports from the Israeli daily newspaper Yedioth Aharonot , and two background commentary articles from the same newspaper; - two essays from the book devarim bego (Explications and Implications; Writings on Jewish Heritage and Renaissance) by Gershom Scholem, Am Oved Publishing House, Tel Aviv, (3rd print 1982) - random selected pieces from the novel menuba nexona (A Perfect Peace), by Amos Oz, Am Oved Publishing House, Tel Aviv,1982 - random selected pieces from the novel lo sam zay in (I don't Give a Damn) by Dahn Ben-Amotz, Bitan Publishing House, Tel Aviv, 1975 The selection attempts to represent a variation of styles, but I am aware that it is too small to be seen as a representation of contemporary Isreali Hebrew. 2. Unlike Stern I consider the PAUL not as a verbal but as a nominal pattern; for reasons- see chapter 1. 3. Note that there need not be a total overlap between valency 2 and transitivity of the verb. Intransitive verbs too can have valency 2, like (i) sit V (xl)Pos (x2)Loc Valency is defined as the number of arguments that a verb can take. Sometimes it is possible to leave an argument unspecified, when the context provides the necessary information. There is, however, a difference between leaving an argument unspecified and reducing an argument. In (ii) below in (a) the Loc is left unspecified, whereas in (b) the Go argument is reduced: (i)a b

John has been sitting reading the telephone directory for an hour now. John drinks heavily.

4. Note that the criterion for deciding when (x2) is Go is a semantic one. Syntactically there is no criterion apart from the use of et. Sometimes it is claimed that not every (x2) which is preceded by et is Go. This is based on examples like (ii)a b

hu haca (Bl) et hatapuah (Go) he halved GM the-apple 'He halved the apple' hu haca (Bl) et harehov (Go/Loc ?) he crossed GM the-street 'He crossed the street'

Here it is asserted that in (iiib), (x2), although preceded by et, is not Go but Loc. However, as the English translation of the verb shows, haca may mean two things, and this is where the difference between the sentences lies. (x2) is in both cases Go, even if in (iiib) the real world entity referred to is a place (note also the difference between FG and Fillmore's Deep Cases in this respect)

164

5. That there is nevertheless a difference can be demonstrated by the use of hifsik 'stopped 1 in a more obviously causative construction than (14a,b). Then, only et + (x2) is possible: (iii)a b

cilcul hapa'amon ringing-of the-bell 'The ringing of the *cilcul hapa'amon ringing-of the-bell

hifsik et hasi'ur. stopped GH the-lesson1 bell stopped the lesson , hifsik lasi'ur /basi'ur— stopped to-the-lesson/in-the-lesson..

Hence, a verb with et+(x2) is not interchangeable in all contexts with a verb + infinitive; therefore the two forms cannot be called synonymous. 6. Stern (1979) also looked at the so called 'bitransitive' or trivalent verbs. He claims that only 210 of the verbs listed in Even Shoshan' dictionary (1967) (totalling 1730 entries) are trivalent. At least one of their arguments (usually (x2) is preceded by et, i.e. is Go. These verbs occur in the following patterns: PAAL 65, FIEL 61 and HIFIL 84. 7. Saad & Bolozky (1984) in fact formulate the derivational regularity of causativization as being based on the intransitive PAAL, i.e. the root can be causativized by the HIFIL only if it is intransitive in PAAL. Our observations, however, do not support this. In the following example (one of many possible) the HIFIL is derived from a +tr PAAL:

(vi) (vii)

8.

yosi Yosi 'Yosi Yosi Yosi 'Yosi

axal (Bl) et kol ha'uga. ate 3p.s.m. GH all the-cake ate all the cake'. heexil et hayeled et kol ha'uga made-eat (B5) 3p.s.m. GM the-child GH all the-cake made the child eat all the cake 1 .

Compare the predicate formation rule suggested in (Dik, 1980)

(viii)

CAUSATIVE FORHATION WITH LATEN input : pred V (xl)— (xn) output: laten-pred V (xO)causer (xl)causee... (xn)

to the rules suggested by Vet: (ix)

causative formation I: input : [ pred V (x2)Go ]Inf (Ax)Ag output:[laten V [pred V (x2)Go]Inf (Axl)Ag]VC (xO)Ag (xl)Go

(x)

causative formation II input : pred V (xl)Ag (x2)go output: [laten[pred V(yl)Ag]inf (Ax2)Go]VC (xl)Ag (x2)Go

(xi)

causative formation III input : Fv perc (xl)Exp (x2)Go output: [laten V Fv perc ]VC (xO)Ag (x2)Go (xl)Rec

9. Causativity in HH can also be expressed by the FIEL pattern. For example (iv)

dan 'icben (B3) et kulam Dan made-angry GH everyone 'Dan made everone angry'

165

(v)

hahom yibes (B3) et kol hacmahim. tbe-beat made-dry GM all the-plants 'The heat dried all the plants'.

The distribution and formation of these causatives productive) is described in Junger (1985a). Here it only that the causative PIEL is always deadjectival. (1977) the old Semitic SIFEL pattern is also express causativity.

(which are not is enough to mention According to Rosen used increasingly to

10. Lexicalized causatives are verbs which could be paraphrased by means of garaa 'caused', but are not used as such, nor is their meaning directly 'cause X ' as with productive causatives. For example (xiii)

maca 'found' himci 'invented' savar 'thought* hisbir 'explained* nigas 'approached' hegiS 'served' balaf 'passed* hehlif 'changed' These HIFIL forms perhaps result from a historical process whereby the HIFIL used to be derived from the PAAL as its causative, but has now acquired an independent lexical status and a slightly different meaning. These HIFIL verbs will be listed in the lexicon separately, as derivatives of the PAAL. 11. The particle Je can have several functions: it can mark Recipient, Beneficiary, Possessor and Experiencer. 12. The passive of (39b) is much more acceptable if the Ag gets a possesive clitic (xii)

hayeled hulbas (B6) al yedei imo (velo 'al yedei the-child was-dressed by mother-bis and-not by haganenet. the-nurse) 'The child was dressed by his mother (and not by the nurse)'.

13. The root x.5.1. 'fail, fall 1 occurs also in the PAAL; xaSal means 'feel, stumbled*. The NIFAL oixSal 'failed', however, is not derived from the PAAL by valency reduction, as it is active intransitive. In my opinion it is derived from the HIFIL of the root, bixSil 'caused to fail', on semantic grounds. The two patterns express a cause-result relation. 14. Reflexivity can also be expressed by the NIFAL, but it is not productive. The only reflexives with NIFAL are (xiv)

Samar 'guarded' - ni§mar 'guarded oneself 3p.s.m. labac 'pushed' - nilhac 'pushed oneself* 3p.s.m. raSam 'note, write down' - nirSam 'enlisted oneself 3p.s.m.

15. In fact g.l.h. (xv)

'shave' + et + Go does exist, as illustrated below:

nezirim budhistiim megalhim (B3) et rosam kol yom. monks Buddhists shave-3p. pl-pres GM head-theirs every day 'Buddhist monks shave their head every day*.

166

This construction, however, is not reflexive, because reflexives are only those verbs which can be paraphrased with the reflexive pronouns 'acmi 'myself, ' acatexa/' acmex 'yourself, 'acmo/'acma 'himself/herself , 'acmenu Ourselves 1 , 'acmexem/'acmexen 'yourselves', 'acmam/'acman 'themselves' 16.

This is in fact evident from the CSS in which the root occurs.

17. Reciprocity can be expressed also by the NIFAL, but it is not productive. In fact the NIFAL is used for reciprocity only in the following case: (xvi)

diber 'talked'

- nidberu 'agreed on something' 3p.pl.m.

18. The denominal or basic HITPAEL forms, illustrated below are also regularities: (xvii)

(xviii)

kalvi hitmaha (B7) af hu behalbanat Kalvi specialized 3p.s.m. also he in-whitewashing-of kaspey hamafia— money-of the-mafia 'Kalvi too, specialized in whitewashing Mafia money—' stalin, aemet be-1953, hitnaged (B7) leheker Stalin, that-died in 1953 opposed 3p.s.m. to-research-of hakibernetika... the-cibernetics 'Stalin, who died in 1953, opposed the research of cybernetics'.

Notes to chapter 5 1. For details on the corpus see the first note to chapter 4. 2. Schwarzwald (1975, 1981b) established the 2.3% by first counting the roots which occur in all seven bioyania in Barkali (1980), which was 11%, and than counterchecking it with Alkalay's dictionaries mi Ion 'ivri anffli Salem Tel Aviv, Hassada Press, 1965 and milon 'ivri Salem RamatGan, Massada Press, 1969. Here she found only 22% of the roots which according to Barkali (1980) would occur in all seven patterns; in total it amounted to 2.3% of the whole list in Barkali (ibid). 3. In Hebrew the two /s/ sounds are spelled with a different letter; this difference is lost in the transcription, which follows the pronounciation and not the Hebrew spelling. This holds for both s.h.t. and s.r.k. (See also 'Notes on the Transcription') 4. Note that the final number of roots is smaller than in Even Shoshan's dictionary, because I omitted some roots which were totally unknown to me. 5. A 'major function' is one that occurs frequently and which is identical to the productive functions as described in chapter 4. The other functions are 'secondary*. 6. The HUFAL pattern is derived from the HIFIL, and not directly from the FIEL. The reason for this is discussed in section 4.2.3.2 .

167

The derivation of the HITPAEL·, whether by a change in qualitative valency (alternatives a, c) or by valency reduction (alternative b) is not relevant to this discussion. 7. Listing all the CSS's and analysing them separately is unnecessary because it will not yield any particularly interesting results. A more important reason for not analysing every single CSS is that the CSS's were formed for the purpose of testing the functions of the biayanim and the derivational relations between them. Therefore, what has to be analysed are these functions and relations; the CSS's are merely a level on which this analysis takes place. 8. 'Resultative' is used to denote the direct result of an action. It is somewhat different from 'inchoative' because inchoative denotes 'acquiring a feature' or 'entering into a state', such as 'becoming cold' for example. Resultative shows a direct connection, mostly causal, between the feature or state described and the action leading to it. Thus 'resultative' is usually the semantic function of a pattern when it is derived from a causative pattern. 9. The possibility of deriving the PIEL and the HIFIL from the PUAL and the HUFAL, has been discarded on the following grounds: a. consistency with the rest of the system b. semantic simplicity. The PUAL and HUFAL in the CSS's in question have the semantic function inchoative (resultative). It rather goes against our intuition to derive the cause from the result (the HIFIL from the PUAL, or the FIEL from the HUFAL·). 10. Schwarzwald (personal communication) suggested that the HITPAEL· is the Resultative of FIEL, but not that the NIFAL is the Resultative of the HIFIL. 11. See note 10. 12. The question which factors determine whether a valency increase of PAAL to form a HIFIL· (c) is possible, is of course an interesting one. In some cases formation of a causative HIFIL with a certain root is impossible because the HIFIL is used with a homophonous root and two HIFIL causatives would cause great confusion. For example the root m.n.t. 'wait* and 'm.t.n. 'moderate' we get him t in 'waited' and therefore 'cause to be moderate* is formed with the PIEL: niteo. This topic is discussed also in chapter 4 section 4.5.3.

References

Alkalay, R. 1965 Milon Ivri- Angli Shalem, Tel Aviv 1969

Kilon Ivri Shalem, Ramat Can

Allerton, D.J. 1982

Valency and the English Verb, London, Academic Press

Ariel, S. 1971 'The system of stem forms in colloquial Israeli Hebrew* in Transactions of the Philological Society, 192-272 Avineri, Y. 1976 heyxal hamiSqalim, Izreel Publishing House, Tel-Aviv Balgur, R.

1965

milot yesod lebeyt hasefer,

Barkali, Sh. 1964 luab haSemot baSalem (Complete list of nominal forms) Rubin Mas, Jerusalem 1980

luab hape'alim haSalem (Complete list of verbs) (revised version: of 1965) Rubin Mas, Jerusalem

Ben Asher, M. 1972a 'The verb patterns: a matter for grammar or the lexicon' in Sauniversita 17(2), 31-34 1972b

'iyuaim betabbir ha'ivrit habadaSa, Hakibbutz Hameuhad Publishing, Tel Aviv

Berman (Aronson), R. 1975a 'The representation of the verb-roots and patterns in the lexicon1 in Sefer Rozen Jerusalem, Council on the teaching of Hebrew, 25-36 1975b

'The morphological realization of syntactic processes in the binyanim system' in Hebrew Computational Linguistics 9, 25-39

1976

On derived and deriving nominals in Modern Hebrew' in P. Cole (ed) Studies in Modern Hebrew Syntax and Semantics, Amsterdam, North Holland Publishing Comp. 57-98

1978a

Modern Hebrew Structure, Tel Aviv, University Publishing Projects

1978b

'kisrey yahasa vehacraxa ba'ivrit hahadaSa' in Orhot 10, 25-33

169

Herman (Aronson), R. 1979a 'Form and function: passives, middles and impersonals in Modern Hebrew in Proceedings of the Third Annual meeting of the Berkeley Linguistic Society, Berkely, vol. 5, 1-27 1979b

'Lexical decomposition and lexical unity in the expression of derived verbal categories in Modern Hebrew 1 in Afroasiatic Linguistics 6, 117-142

1982a

'Dative marking of the afectee role: data from Modern Hebrew' in Hebrew Annual Review, Department of Judaic and Near Eastern Languages and Literatures, Ohio State University

1982b

'Verb pattern alternation: on the interface of morphology syntax and semantics in Hebrew Child Language' in Journal of Child Language 9/1, 169-191

1982c

On the nature of Oblique' objects in bitransitive constructions' in Lingua 56, 101-125

1983

'Establishing a schema: children's construals of verb-tense marking* in Language Sciences 5/1,

Berman, R & Υ. Sagi 1981 '*al darkey tecurat hamilim vehidusan bagil haca'ir' in Hebrew Computational Linguistics 18, 32-61 Blau, Y. 1967 dikduk 'irri Sitati, Jerusalem, Hamaxon Hayisraeli Lehaskala Bextav 1972

yesodot hatabbir Jerusalem, Hamaxon Hayisraeli Lehaskala Bextav

Bolkestein, A.M. 1983a 'The role of discourse in syntax: evidence from the Latin Nominativus cum Infinitive' in K. Ehlich & H. van Riemsdijk (eds) Connectedness in Sentence, Discourse and Text, Department of Language and Literature, Tilburg University 1983b

'Genitive and dative possessors in Latin' in Dik (ed) 1983, pp. 55-91

1985

'Discourse and case-marking: three place predicates in Latin' in Ch. Touartier (ed) Syntax et Latin, Aix en Provence: Universite de Provence

Bolkestein, N.A., C. de Groot & J. L. Mackenzie (eds) 1981 Predication and Expression in Functional Grammar London, Academic Press Bolozky, S. 1972 On morphological and phonetic constraints in Modern Hebrew' in Hebrew Computational Linguistics Buletin 6,

170

Bolozky, S. 1978 'Word formation strategies in the Hebrew verb system: denominative verbs' in Afroasiatic Linguistics 5, 111-136 1983

'Strategies of Modern Hebrew Word Formation', paper based on a lecture given at the North American Conference on Afroasiatic Linguistics and the American Oriental Society, March 29, 1982

Boogart, P.C. uit den, (ed) 1975 Word Frequency in Spoken Dutch, Oosthoek, Scheltema & Holkema, Amsterdam Brown, D.R. 1984 'Terms and term operators', paper, SIL Nairobi 1985

'Term operators' in Bolkestein, De Groot & Mackenzie (eds) 1985, 127-145

Cole, P. (ed) 1976a Studies in Modern Hebrew Syntax and Semantics, Amsterdam, North Holland Publishing Comp. 1976b

causative construction in Modern Hebrew: theoretical implications', in Cole (ed), 99-128

Comrie, B. 1981 Language Universals and Linguistic typology; Synatx and Semantics, Oxford, Basil Blakwell Ch. 8

Dik, S.C. 1978 Functional Grammar Amsterdam, North Holland Publishing Comp (3rd ed. Foris, Dordrecht) 1979a

Seventeen Sentences: basic principles and application of Functional Grammar, Publication of the Institute for General Linguistics, University of Amsterdam

1979b

'Functionele Morfologie' T. Hoekstra & H. van der Hülst (eds) Morfologie in Nederland, GLOT-special Leiden, 72-100

1980

Studies in Functional Grammar, London/ New York, Academic Press

1983a

On the status of verbal reflexives' in Communication and Cognition 6, 39-63

1983b

(ea)Advances in Functional Grammar, Dordrecht, Foris

1985

'Valentie en valentie operaties in Functionele Grammatika 1 in Interdisciplinair Tijdschrift voor Taal- en Tekstwenschap, 5/2, Foris, Dordrecht, pp. 95-114

1987

The Theory of Functional Grammar (in preparation)

171

Dik, S.C. & J. Gvozdanovic 1981 'Subject and Object in Servo-Croatian' in Hoekstra et al (eds) 21-39 Driver, S.R. 1879 A Treatise on the Use of the Teases in Hebrew, Oxford, Clarendon Press Even-Shoshan, A. 1967 Milon Hadash, Kiryat Sefer Publ., Jerusalem Fillmore, C. 1968a 'The case for case' in Bach & Harms (eds) Universals in Linguistic Theory, Holt, Reinhart and Winston, 1-90 Frajzyngier, Z. 1982 'Indefinite agent, passive and impersonal passive: a functional study' in Lingua 58, 267-290 Gesenius 1910 Hebrew Grammar ed. by E. Kauzch, revised by A.E. Cowley, Oxford, Clarendon Press Gordon, A. 1982 'The development of the participle in Biblical, Mishnaic and Modern Hebrew', in Afroasiatic Linguistics, 8/3, 121-179 Greenberg, J. 1950 'The patterning of root morphemes in Semitic' in Word 6, 162-182 Goshen-Gottstein M., Z. Livne & Sh. Sapan 1967 hadikduk ba'ivri hasimuSi, Jerusalem, Shoken Hannay, H.

1985

English Existentials in Functional Grammar, Functional Grammar Series 3, Dordrecht: Foris

Hoekstra, T., H. van der Hülst & H. Hoortgat (eds) 1981 Perspectives on Functional Grammar, Dordrecht, Foris Hopper, P.J. & S. Thompson, 1980 'Transitivity in grammar and discourse' in Language 56/2, 251-299 Junger, J. 1983 On the problem of the passive in constructions of perception and cognition in Modern Hebrew', in: H. Bennis & W.U.S. van Lessen Kloeke (eds) Linguistics in the Netherlands Foris, Dordrecht, pp. 115-123 1985a

'Morphological causatives in Modern Hebrew', in: A.M. Bolkestein, C. de Groot & J.L. Mackenzie (eds) Predicates and Terms in Functional Grammar, Foris, Dordrecht, pp. 235-257

1985b

'Valentie reductie in het Modern Hebreeuws', in: Tijdscbrift voor Taal- en Tekst Vetenscbap nr. 2 pp. 141-160

172

Junger, J. 1986 'Agentless passives in Modern Hebrew', to appear in: Acta Linguistica Hafniensia Kimhi, R. D.

1847

Sefer Hashorasbim, Berlin

Kwee, T.L. 1979 'A68-FG(3); Simon Dik's functionele grammatika geschreven in algol 68 versie nr. 03' Publication nr. 23 of the Institute for General Linguistics, University of Amsterdam 1981

'In search of an appropriate relative clause' in Hoekstra et al (eds) (1981)

McCarthy, J.J. 1981 Ά prosodic theory of non-concatenative morphology 1 in Linguistic Inquiry 12/3, 373-417 Mentcher, E. 1983 'Expressions of causativity in English and Hebrew' in Language Across Culture Proceedings of IRAA Symposium, 155-169 Mitchell, T.F. 1978 'Educated spoken Arabic in Egypt and the Levant, with special reference to participle and tense' in Journal of Linguistics 14, 227-258

Nir, R. 1979 'hamivne hasemanti ael Smot ha'ecem hamurkavim Seba'ivrit hahadaSa* in Hebrew Computational Linguistics 15, 9-17 Nuyts, J. 1985 On Formalism, Functionalism and Mentalism: towards a Functional Procedural Grammar' in Nuyts (ed) (1985), 113-168 1985

(ed) Antwerp Studies in Functional Grammar, Antwerp Papers in Linguistics nr. 39

Ornan, U. 1969 'Review of the Hebrew verb system', Mimeograph, Hebrew, University of Jerusalem 1971a

'Verb patterns and stems, inflections and declensions' in Ha'universita 2, 15-22

1971b

'binyanim ubsisim, netiyot ugzarot 1 in Ha'universita 16, 15-22

1971c

'New approach to the description of the verb and its teaching, an example from the declension of the benoni' in Sh. Kodesh (ed) Kamrat memorial volume, Jerusalem, The Council for the Teaching of Hebrew, pp. 32-44

1978

'netiyat hapo'al keycad? 1 in Orttot 1978, 47-68

1979

On the meaning of the binyanim', in minha lekodesh, Jerusalem,The Council for the Teaching of Hebrew, pp. 11- 19

173

Risselada, R. 1987 'Voice in Ancient Greek: reflexives and passives' Paper, University of Amsterdam Rosen, H. 1977 Contemporary Hebrew, Mouton, The Hague Rubinstein, E. 1971 The verbal phrase: studies in contemporary syntax (in Hebrew), Tel Aviv, Hakkibuts Hameubad Publishing Saad G. & Sh. Bolozky, 1980 'Theoretical implications of the morphological causativization in Arabic and Hebrew', Paper, University of Hassachusets 1984

'Causativization and transitivization in Arabic and Modern Hebrew 1 in Afroasiatic Linguistics 9/2, 101-110

Sasson, B. 1976 torat bapo'al, Tel-Aviv, Yavneh Publishing House (ch. 6) Schwarzwald, 0. 1974 'Roots, stems and the structure of morphemes' in lesbonenu 38, 131-136 1975

'Comments on roots-pattens relations in the Hebrew lexicon* in Hebrew Computational Linguistics 9, pp. 47-59 (in Hebrew)

1977

'The representation of the weak verb in the Modern Hebrew lexicon', in Hebrew Computational Linguistics 12, 25-36 (in Hebrew)

1981a

'Grammaticality in Modern Hebrew* in loteroational Journal of Middle East Studies, 13, 11-19

1981b

'Frequency factors as determinants in the Binyanim meaning* in Hebrew Studies, 22, 131-137

1981c

Grammar and Reality in the Hebrew Verb Ramat Gan: Bar Ilan University Press

Shalgi, M. 1976 dikduk Sem ba'ecem: nilcudo umaSma'uyotav, Tel-Aviv, Or Am Publishing House

Shibatani, M. 1976 'The grammar of causative constructions: a conspectus' in M. Shibatani (ed.) Syntax & Semantics 6: on the Grammar of Causative Constructions, London, New-York, Academic Press Stern, N. 1981 'et-verbs in Israeli Hebrew' in Hebrew Computational Linguistics 15, 28-55 1981

'Trivalent verbs with an et and other arguments' in Hebrew Computational Linguistics 17, 46-61

174

Vester, E. 1983 Instrumeat and Manner Expressions in Latin, PhD Dissertation, University of Amsterdam Waiden (Peres), Z. 1982 The root of roots: children's construction of word formation processes in Hebrew, PhD dissertation, Harvard University Werner, F. 1982 'haefSaruyot haSonot Sei tecurat hamilim ba'ivrit bat zmanenu' in Proceedings of the 8th World Congress on Jewish Studies, (1981), 57-62

Author index

Alcalay, R. 102 Allerton, O. J. 62 Ariel , S. 21£f. f 29 Avineri, Y. 50 Balgur, R. 81, 103 Barkali, S. 71, 72, lOOff. Ben-Asher, N. 23 Berman, R. 19ff., 25, 27, 44, 45, 47, 53, 6 4 f f . , 69, 74, 84 88, 90, 98, 107, 155 Blau, Y. 19, 95 Bolkestein, A.M. 31, 35, 156, 158 Bolozky, S. 43, 46, 64, 69, 71, 74, 88, 107, 155, 157 Brown, D.R. 35 Cole, P. 74, 76ff. Comrie, B. 74 Dik, S.C. 32f£., 34, 38, 60, 63, 74, 80, 86, 88, 141, 156 Driver, S.R. 13 Even-Shoshan, Sh. 72, 89, 91, lOOff. Fillmore, C. 36 Frajyzyngier, Z. 84 Gesenius 13ff., 90, 98 Gordon, A. 16 Goshen-Gottstein, K. 19, 23 Greenberg, J. 47 Gvozdanivic, J. 88 Hopper, C. 64 Junger, J. 35, 80, 85, 96, 98, 157 Kimhi, R. D. 13, 16f£. Kwee, T.L. 33 Mackenzie, J.L. 63 McCarthy, J.J. 2 3 f f . , 48 Menteher, E. 74 Mitchell, T.F. 12 Moutaouakil, A. 75, 156 Nir, R. 51 Oman, U. 17ff., 98 Risselada, R. 156 Rosen, H. 23, 51, 66 Rubinstein, E. 23 Saad, G. 62, 65, 69, 74 Sagi, Y. 25, 47, 64, 90 Sasson, B. 13, 23, 27, 29 Schwarzwald, O. 18ff., 25, 26, 43, 44 47, 53, 98, 102, 155 Shalgi, M. 50 Shibatani, M. 74 Stern, N. 65£f., 87 Thompson, S. 64 Van Schaaik, G. 63, 96 Vet, C. 74, 96, 156 Waiden, Z. 25, 47, 82 Werner, F. 48

Subject index

Arabic

12, 23, 24, 43 causativity 75 Basic predicate 34, 35, 42 as lexical entry 45-48 Basic predicate-frames 34 Basic terms 35 Benoni 16, 27 By-phrase ( 'al yedei') 84-86 in pseudo-passive constructions 84-86 Biblical Hebrew verbal system 16 Binyanim basic: definition 70 basic: criteria for establishing 52, 70 in CSS's 122 in the lexicon 71, 72 Binyanim (general) 12ff. basic and derived 70ff. derived 72, 73 functions 64 in pseudo-passive constructions 86 generalized predicate-schemes 64 as organizing principle of MH verbal system 152 productive functions 64 productive processes 64 representation in lexicon 51ff. Binyanim system 13ff., lexicalist view 17 non-concatenative morphology 23 traditional view 13 TGG view 19 Binyania: syntactic functions 19ff. causativity 21, 7 3 f f . inchoativeness 21, 86 ingression 21, 86 pseudo-passive 83ff. reflexivity 20, 86 voice 20, 83ff. Causative construction 7 4 f f . derivation 81 Causer and Causee 76 in the 'passive' 78 Causative State of Affairs 79, 80 Causativity 7 4 f f . definition 74 predicate formation 81 morphological (HIFIL) 85ff. Caused State of Affairs 78, 79 Causee 76ff. semantic function 76-80

177

Child language 19ff. bioyaaia system 19, 25, 26 miSqalia system 25 Configurational Sub-Systems 40, 9 7 f f . complete list 108 in lexicon 54 procedure for establishing lOOff. binyanint in CSS's 54 bioyaoia functions 98

definition 99 size 117 size and productivity 117ff., 152 valency change 97 consistency in valency 120 consistency of quantitative valency of multi-functional binyanim 119ff. consistency of synactic-semantic function of binyanim 119ff. derivational relations 121 derivational relations between binyanim 121, 153ff. explaining gaps in verbal system 154 large CSS's 117 medium size CSS's 118 numerical distribution of roots in CSS's 106, 107 regularity in biayaoim functions in CSS's 96, 119ff. small CSS's 118 valency changes 62, 121ff. Derivation of HIFIL causatives 81 of HITPAEL detransitive 92 of NIFAL, PUAL, HUFAL pseudo-passive 83-88 Derived binyanim 72 Detransitivization 83ff. Detransitivized predicate (generic) 83 Dictionary 43 Double reduction of valency 93 predicate formation rule (B6 ( p p ) ) 94 Experiencer in causative constructions 78-79 First argument reduction 83 Fund 34 Generalized predicate-scheme 40 in lexicon 53 Goal in causative constructions 76 HIFIL 15ff., 20 basic and derived 75 causative, gaps 81 causative, predicate formation rule 81 denominal 144 denominal in CSS's 144 derived from FIEL 143 double reduction of valency 93 expressing morphological causative 75 HIFIL causative 20 derivation 81 psychological reality 75, 81 as lexicalized causative 81 productivity 75

178

HISTAFEL 29 HITPAEL 17, 19, 20

as basic binyan 71 basic and derived 73 derived by qualitative valency change 140 derived from PAAL/PIEL/HIFIL 130 in child language 90 inchoative, with NIFAL (i) in same CSS 136 inchoative 90, 134 inchoative, text frequency 91 inchoative-resultative 92, 134 inchoative-resultative, derivation 136 intransitive 90-91, 132 intransitive, text frequency 91 intransitive in CSS's 134 reciprocal, derived from PAAL 140 reciprocal, derived from PIEL 140 reciprocal 90 reciprocal, in CSS's 139 reciprocal, text frequency 91 reciprocal, derivation 90 reflexive, derivation 89 reflexive 89 reflexive in CSS's 136 reflexive, predicate formation rule 89 reflexive, text frequency 91 HITPUAL 29-30 HUFAL 17, 19, 83 always derived 72 double valency reduction 93 derivation 83 Intransitive 90, 91 Lexicon 33ff., 43ff. approaches to 43, 44 consisting of CSS's 55ff. CSS's as entries 55 FG view 48 organization 53 M FOAL 27 Misqalim 17, 25, 26, 50 in lexicon 50 system, regularity 50 Hishnaic Hebrew 16 Horphological patterns nominal and verbal 43 HUFAL 27 NIFAL 14, 83, 86, 124 basic and derived 73 derived from HIFIL 126 inchoative-resultative of RIFIL (c) 126 NIFAL· formation 86 HITPAEL 28 Non-productive processes 152 Non-productive regularities 37 PAAL 14, 17, 19, 27, 124, 128 source for B7(ref) 137 the unmarked binyan 70 PAUL 27

179

Passive

8 3 f £ . , 121-125

Subject assignment 81 argument reduction 81, 93 FIEL 15, 18, 19 as causative 80 basic and derived 73 derived from PAAL by qualitative valency change 147 source for HITPAEL (ref) 137 POEL 17 Predicate formation rules 36 as valency change 64 Predicate-frame 35, 37, 39 Predicate-frames 35 Predicate-scheme 40, 63 in lexicon 51 applicability of predicate-frames 154 Predicates basic 38 derived 37, 38 Productive processes 152, 153 Productive rule predicate formation 37 Productivity 37, 152, 155 qualitative criterion vs. size of CSS's 155 qualitative criterion 37 quantitative criterion 37 Prototypical functions binyanim 53 of binyanim in lexicon 53 Prototypical relations binyanim 53 Pseudo-passive 128 NIFAL 8 3 f f . , 124 HUFAL· 8 3 f f . , 127ff. PUAL· 8 3 f f . , 127ff.

predicate formation 86 of causative constructions 78 PUAL 15, 19, 8 3 f f . , 86, 127ff. always derived 72 PUAL formation rule 86 Qualitative changes of valency in CSS: NIFAL 124 Qualitative valency change, PAAL -> PIEL 146 Qualitative valency change CSS's, HITPAEL 140, 141 Reflexivity 20 in MH 88 Regularity 38 Root 39, 45, 46 in lexicon 48 secondary (extracted) 49, 106 Root-binyanim distribution 107

180

Roots

42ff. homophonous 106 in predicate-frames 48 notation in sources 103 notation in sources (differences) 103-106 number of, in sources 105 representation in lexicon 48 SAFEL (SIFEL) 28-29 Second argument reduction 88 Semantic function of (x2) 68, 69 (x2) in causative constructions 7 6 f f . Semitic 13, 23, 45, 47 State of Affairs 35 Stem 39 Subj assignment 69 Subj function assignment, passives 82 SUFAL 29 Syntactic-semantic relations between the binyanim 22 The bioyania system 12ff. traditional view 13 transformational generative view 19 lexicalist view 17 FG view 25 Transitivity 64ff. in the binyanim system 64 present FG definition 62 quantitative valency 69 semantic definition 65 syntactic definition 65 Transitivization 73ff. valency increase 73 Valency 40ff. change, as predicate formation rules 62 general 62 of predicate 62 qualitative 62 quantitative 62 related to transitivity 64ff. Valency change as predicate formation 64 in CSS's 120ff. reduction 83ff. increase 73ff. Valency increase 73, 141ff. PAAL· -> PIEL in CSS's 145 PAAL -> HIFIL·

141

causativization 7 3 f f . transitivization 7 3 f f . Valency reduction 68, 82 1st argument: pseudo passive 83 2nd argument 88 double reduction (rule) 92 in CSS: HIFIL -> HUFAL 127 in CSS: PAAL -> NIFAL 123 in CSS: PIEL -> PUAL, HIFIL -> HUFAL

127

181

Verbal system Arabic 64 Hebrew 64 representation in the lexicon 47