Unity in Diversity: Papers Presented to Simon C. Dik on his 50th Birthday 9783110847420, 9783110133530

Table of contents :
Preface
The role of quotations in Andean discourse
Unreportable linguistic entities in Functional Grammar
Issues in Functional Discourse Analysis
Intransitivity, transitivity and control in Lillooet Salish
The life of language
Auxiliary operator Inf and constructions with an infinitive in Hungarian
A note on semantic functions and adpositional term predicates
Politeness and mitigation in Spanish: a morpho-pragmatic analysis
Normative and analytical perspectives on interviewing techniques
Quechua mu and the perspective of the speaker
The discourse structure of job interviews
Translation studies: developments and perspectives in teaching and research
The development of cases and adpositions in Latin
Adverbs in Papuan languages. Or: “Where have all the adverbs gone?”
Functional prepositions
On the internal structure of the Noun Phrase in Bahasa Indonesia
Deverbal nominalization in Sranan: a search for regularities
Aktionsart, aspect and duration adverbials
Some remarks on direct quotation in Kombai

Citation preview

Unity in Diversity

Harm Pinkster, Inge Genee (eds)

Papers presented to Simon C. Dik on his 50th birthday

Ψ 1990

FORIS PUBLICATIONS Dordrecht - Holland/Providence Rl - U.S.A.

Published by: Foris Publications Holland P.O. Box 5 0 9 3 3 0 0 A M Dordrecht, the Netherlands Distributor for the U.S.A. and Canada: Foris Publications USA, Inc. P.O. Box 5 9 0 4 Providence Rl 0 2 9 0 3 U.S.A.

CIP-DATA

ISBN 9 0 6 7 6 5 511 2 © 1 9 9 0 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.

Contents

Preface

vii

Willem F.H. Adelaar The role of quotations in Andean discourse

1

Α. Machtelt Bolkestein Unreportable linguistic entities in Functional Grammar

13

Teun A. van Dijk Issues in Functional Discourse Analysis

27

Jan P. van Eijk Intransitivity, transitivity and control in Lillooet Salish

47

Dorothea Franck The life of language

65

Casper de Groot Auxiliary operator Inf and constructions with an infinitive in Hungarian

73

Mike Hannay A note on semantic functions and adpositional term predicates

93

Henk Haverkate Politeness and mitigation in Spanish: a morpho-pragmatic analysis Martha L. Komter The discourse structure of job interviews Hanneke Houtkoop-Steenstra

107 165

Normative and analytical perspectives on interviewing techniques

133

Simon Quechua vanmu de and Kerke theand perspective Pieter Muysken of the speaker

151

vi Kitty Μ. van Leuven-Zwart Translation studies: developments and perspectives in teaching and research

183

Harm Pinkster The development of cases and adpositions in Latin

195

Ger P. Reesink Adverbs in Papuan languages. Or: "Where have all the adverbs gone?"

211

Henk van Riemsdijk Functional prepositions

229

Sie Ing Djiang On the internal structure of the Noun Phrase in Bahasa Indonesia

243

Norval Smith DeverbalJ.nominalization in Sranan: a search for regularities Lourens de Vries Some remarks on direct quotation in Kombai Co Vet Aktionsart, aspect and duration adverbials

265 291 279

Preface

Simon Cornells Dik was born in September 1940. He entered the University of Amsterdam at the age of eighteen, to study Classics and Linguistics. He was at once taken on as a research assistant by the then Professor of Linguistics, Anton Reichling. After his appointment as Professor of Linguistics in 1969, he created (together with his colleague Ben Tervoort) a truly multi-disciplinary department. He has been active in all sorts of administrative duties. Most importantly, he is the author of a large number of books and articles on a broad range of subjects, from morphology to conversational analysis, from Dutch to non-IndoEuropean languages, both descriptive and theoretical, diachronic and synchronic. Among these publications his work on Functional Grammar is a source of inspiration for an active group of "functional grammarians", both in the Netherlands and abroad. The diversity of his scholarly interest is reflected in the range of Ph.D. dissertations he has supervised or co-supervised, 26 to date and several other dissertations underway. On the occasion of his 50th birthday, which for Simon Dik is also the beginning of a part-time professorship (to have more time for his own research), eighteen of his promotae and promoti have united in writing a collection of papers for their promotor, with gratitude and respect. We had the privilege of coordinating and editing the work.

Harm Pinkster Inge Genee

The role of quotations in Andean discourse

William F.H. Adelaar Department of Comparative Linguistics State University, Leiden

0.

Introduction

The rendering of conversations in a narrative account normally involves the following ingredients: (1) an expression (most commonly a verb) referring to the act of communication and (2) an expression referring to the message communicated. The latter either takes the shape of a grammatically embedded object clause, which may be marked by the presence of a complementizing element, or it is reproduced in its original form, that is, as a quotation. In a number of Western European languages, the two construction types coexist and the selection of either one can be a matter of style or the preference of the speaker. For instance, in Dutch we find competing options, such as the following: (la)

Hij zei tegen mij: "Ga weg!". 'He said to me: "Go away!".'

(lb) Hij zei dat ik weg moest gaan. 'He told me that I should go away.'

2

Willem F.H. Adelaar

(2a) Hij vroeg aan mij: "Slaap je?". 'He asked me: "Are you asleep?".' (2b) Hij vroeg me of ik sliep. 'He asked me whether I was asleep.' Several native languages of the Andean region also have constructions involving embedded object clauses and direct quotations, but the decision as to which type of construction must be used in each particular instance is not a matter of free choice. The present contribution is intended to summarize a number of relevant facts concerning the expression of message complements in the Andean languages Quechua and Araucanian. It also addresses the consequences of a grammatically conditioned division between quotation complements and clausal object complements.

1. Message complements in Quechua In Quechua, the equivalent of the verb "to say", ni- (ni- in some dialects), can be accompanied by a quotation but not by an embedded object clause that expresses the content of a message. At the same time, rti- is the only verb that can occur in direct syntactic contact with a section of quoted speech. Verbs other than ni- that refer to some kind of communication ("to tell", "to warn", "to ask", "to call", "to argue", etc.) must be used in a periphrastic construction involving a gerund-like form of the verb ni- preceded by a direct quotation, whenever the content of the message is to be overtly expressed. This gerundlike form, nispa (nispa) closes any section of quoted speech and establishes a necessary link between such a section and verbs of communication other than ni-. Impressionistically, it acts as the vocal counterpart of a colon in English. The examples (3) and (4) illustrate the above. (For an explanation of the glosses see the note at the end of this article.) (3)

"uku wayqu-pi allpa-kuna-qa ancha sumaq-mi"ni-n-ku. deep valley-in fields-PLUR-TOP very fine-ASS say-3s-PLURs 'They say: "Down in the valley the fields are excellent."'

The role of quotations in Andean Discourse (4)

"kay-man chay makawisa-kta apa-mu-chun" fii-spa this-ALL that Macauisa-ACC take-CIS-3sIMP say-GER

kacha-rka-n. order-PAST-3s. 'They gave an order saying: "Let him bring Macauisa here!"' OR: 'They gave the order that he should bring Macauisa to them.' The verb ni- itself is frequently found in the same type of construction. In it, nioccupies the place of the higher verb and is accompanied by its own gerund. This can be seen in the following example (5). (5)

"lliw allpa-kuna-m qam-kuna-paq ka-nqa" all field-PLUR-ASSyou-PLUR-BEN be-3sFUT

ni-spa-s-iki kanan-qa ni-wa-chka-nchik. say-GER-HRS-APP nowadays-TOP say-3s4o-DUR-3s4o. 'As you should know, they now reportedly say to us: "All the fields will be yours".' OR: 'As you should know, they now reportedly say to us that all the fields will be ours.' In contrast, several verbs referring to perception or knowledge ("to hear", "to see", "to know", "to remember") or reflection ("to speak about", "to discuss") take object clauses as their complements, as is illustrated in (6), (7) and (8). (6)

senor-ni-nchis yacha-n-mi chay alma-kuna-q moroso Lord-CONN-4p know-3s-ASS those deceased-PLUR-GEN slow payer ka-sqa-n-ku-ta. be-NOM-3ps-PLUR-ACC 'Our Lord knows that those deceased people have been slow payers.'

(7)

hina riku-y-ku sikwani uray-man chinka-ya-pu-q-ta. then see-ls-PLUR Sicuani below-ALL disappear-DIR-NORET-NOMACC 'Then we saw (the airplane) disappear downward towards Sicuani.'

4

Willem F.H. Adelaar

(8)

fta-m-ari wakin-ni-n kapitulu-pi as-lla-kta already-ASS-APP other-CONN-3p chapter-LOC a little-RESTR-ACC yuri-mu-ska-n-ta-pas rima-rka-nchik. be bora-CIS-NOM-3ps-ACC-INCL speak-PAST-4s 'As you may remember, we have already spoken a little of the way he came to be born in a previous chapter.'

In sentences (6), (7) and (8), the presence of nominalizers such as -sqa (-ska in the variety of Quechua represented in (8)) or -q marks the clauses referring to the fact known, the event perceived or the topic reflected upon as a potential complement. The three constructions introduced in this section can be formally summarized as follows: (I)

"QUOTATION" ni-

(Ila) (Ilb)

"QUOTATION" nispa V"QUOTATION" nispa ni-

(III)

OBJECT CLAUSE (NOM) V-

2. The meaning of "to say" The outline given in section 1. suggests a division between a class of communication verbs consisting of ni- and a number of semantic equivalents of ni- (cases I and II) on one hand and a class of verbs of knowledge, perception and reflection on the other (case III). The reality, however, is more complex, and this has to do with the unusually extensive semantic domain covered by ni-. The verb ni- can refer to all kinds of mental activity in which no actual talking is involved. (The expression "inner speech" used by De Vries, this volume, is quite appropriate for referring to the content of such activities.) Among the semantic readings potentially assigned to ni- we find such interpretations as "to consider", "to want" and "to intend", along with cases of actual speaking ("to call", "to answer"). Variation in the choice of lexical material is not a stylistic requirement for speakers of Quechua, a fact which may explain in part the enormous frequency of ni- and other basic vocabulary (verbs in particular) in Quechua texts. In short, the meaning of ni- to a large extent extends beyond that of a verb of communication. And, as we have seen before, the verb ni- demands a message complement in directly quoted speech (with or without the intervention of nispa).

The role of quotations in Andean Discourse

5

The above combination of facts compels speakers to use quotations in sentences which do not refer to actual speaking in order to render the content of a mental activity. The following examples illustrate such uses. (9)

chay-si kiki-lla-n-tak uk wata chika nunu-n-wan therefore-HRS self-RESTR-3p-SEQ one year about breast-3p-INSTR

kawsa-chi-rka-n "pi-p churi-n-ch kay-ka" iii-spa. live-CAUS-PAST-3s who-GEN son-3P-CONJ this-TOP say-GER 'So she herself reportedly kept (the child) alive with her own breast for about one year while saying: "Whose son could this be?'" OR: '...wondering whose son it could be.' (10) "fiuka-kta-pas muna-wa-nka" iii-spa amu-rka-n-ku. me-ACC-INCL want-lo-3sFUT say-GER come-PAST-3s-PLUR 'They came while saying: "She will love me as well.'" OR: 'They came with the expectation of also being loved by her.' (11) "mana-chu-s ri-sak ri-sak-chu-s" fii-k ina-s not-INT-HRS go-lsFUTgo-lsFUT-INT-HRS say-NOM like-HRS ayni-kacha-rka-n. answer-HESIT-PAST-3s. 'He hesitated to give his answer as if he were saying: "Am I not supposed to go or am I supposed to go?'" OR: 'He hesitated to give his answer as if he were in doubt whether he should go or not.' In none of these examples does the verb ni- "to say" refer to actual speaking. The uses illustrated call for English translations such as "to wonder", "to expect", "to doubt." However, direct quotations are required as complements of ni- in exactly the form they would take if the referent of the subject of ni- were actually to pronounce his thoughts.

3. Nonsubject encoding and the verb "to say" In Quechua verbs, the personal reference system obligatorily encodes not only subjects, but also human nonsubjects if these correspond to the speaker, the addressee or both. The semantic function (as defined by Dik 1980: 3) of an encoded nonsubject is not a priori established and depends on the meaning of the verb in question. With the verb ni- it may be an addressee (the person to

6

Willem F.H. Adelaar

whom is spoken), a person about whom is spoken or even a person in relation to whom something is said. The encoding of nonsubjects is also normally respected in those cases where ni- does not refer to actual speaking. This usage further strengthens the impression of the subject being engaged in a conversation, even though nothing but thoughts or wishes is reproduced. The following examples illustrate this: (12) ancha if

yaku-wan atipa-pti-y-ka "yaya-nchik-mi water-INSTR win-SUB-ls-TOP Lord-4p-ASS

atipa-n" win-3s

ni-spa-m ni-wa-nki. say-GER-ASS say-lo-2s 'If I should be victorious with water, you will say about me: "Our Lord is winning."' OR: 'If I should be victorious with water, you will consider me, your Lord, to be the winner.' (13) "miku-y-ni-yki-kta usu-chi-sak" fii-pti-yki-pas eat-NOM-CONN-2p-ACC be spoiled-CAUS-lsFUTsay-SUB-2s-INCL chay kapari-y-ni-yki-kta uyari-spa-m ancha utka-lla that screech-NOM-CONN-2p-ACC hear-GER-ASS very rapidRESTR karku-su-nki. chase-3s2o-3s2o 'Whenever you will say (in relation to someone): "I shall spoil your food", he will hear your screeching and chase you away.' OR: 'If you intend to spoil somebody's food, he will hear your screeching and chase you away.' See (5) for another example of nonsubject encoding with the verb ni·.

4.

Quotations within quotations

A section of quoted speech acting as the message complement of ni- "to say" may itself contain an instance of ni- accompanied by a quotation. Quotations included within quotations are relatively frequent as a result of the necessity of expressing the respective messages by means of quoted speech. The resulting constructions may acquire a considerable complexity. For each instance of nithe speaker must seek the perspective of the subject corresponding to it and fill in the required affixes of personal reference accordingly.

The role of quotations in Andean Discourse

7

(14) chay-si yaya-n-ta-ka "yaya, kay-pi-m uk so-HRS father-3p-ACC-TOP father this-LOC-ASS one wakcha [yaya-yki-kta alli-ya-chi-sak] iii-mu-wa-n" poor man father-2p-ACCgood-become-CAUS-lsFUT say-CIS-lo-3s iii-spa willa-rka-n. say-GER tell-PAST-3s 'So she reportedly told her father: "Father, a poor man here says to me: [I will cure your father].'" OR: '..."Father, a poor man here says to me that he will cure you."' In (14), a girl reports a conversation for which she has to adopt the perspective of an unexpected visitor whose words she transmits literally. The choice of the right personal reference affixes to express this perspective is crucial for a correct interpretation. A further illustration is (15). (15) chay-si chay fiawpa-k churi-n-pa kusa-n-ka kay so-HRS that precede-NOM child-3p-GEN husband-3p-TOP this simi-kta uyari-spa-s ancha pifla-rka-n news-ACC hear-GER-HRS very become angry-PAST-3s "ima-pak-mi [chay uk wakcha-wan tinki-chi-sak] what-BEN-ASS that one poor man-INSTR be united-CAUS-lsFUT iii-spa fii-n fiuka say-GER say-3s me

chika kapak-pak such rich man-GEN

kunada-y-ta" iii-spa. sister-in-law-lp-ACC say-GER 'Then the husband of his eldest child reportedly became very angry when he heard the news, and said: "For what reason does he say: [I will marry her to that one poor fellow], she being my sister-in-law and I being such a rich man?'" OR: '..."Why does he intend to marry her to that one poor fellow, she being my sister-in-law and I being such a rich man?'"

8 5.

Willem F.H. Adelaar Hearsay evidentials

Quechua sentences normally contain an indication of the source from which the information transmitted was obtained. The affixes that serve this purpose operate at the sentence level and are commonly referred to as evidentials or validators. The set of evidentials includes an affix -s/-si that indicates hearsay and second-hand information (translated 'reportedly' in the examples given so far). The availability of a hearsay evidential offers an alternative for the outermost frame of communication reference and produces the effect of mitigating the complexity of sentences involving quotations within quotations. One must observe, however, that the use of a hearsay evidential does not provide any clue for identifying the person who supplied the information. Hearsay evidentials help the speaker disclaim responsibility for what he says ("I only heard it said, I did not see it myself). If a speaker wants to make clear that a known person supplied information that he is transmitting, he will use niinstead of the hearsay evidential. A good illustration of the interaction between a hearsay evidential and the ni- + quotation construction is supplied by example sentence (5), repeated here for the sake of clarity as (16). (16) "lliw allpa-kuna-m qam-kuna-paq ka-nqa" all field-PLUR-ASS you-PLUR-BEN be-3sFUT ni-spa-s-iki kanan-qa ni-wa-chka-nchik. say-GER-HRS-APP nowadays-TOP say-3s4o-DUR-3s4o 'As you should know, they now reportedly say to us that all the fields will be ours.' It can be argued that, in (16), the hearsay evidential -s- in ni-spa-s-iki is in a certain sense equivalent to an additional ni-spa ni-n-ku ("saying, they say") frame.

6.

Switch-reference

The presentation of dialogues in Quechua texts depends to a large extent on a phenomenon commonly referred to as switch-reference in linguistic literature (see, for instance, Weber 1989: 298). Switch-reference implies that subordinate verbs are marked for having same or different subjects in relation to a hierarchically higher verb in the same sentence. Most Quechua dialects distinguish two affixes which have the function of subordinating verbs to other verbs. The suffix -spa, glossed 'GER' (gerund) in the example sentences, implies

The role of quotations in Andean Discourse

9

subordination to a verb with the same subject; the suffix -pti-, glossed 'SUB' (subordinator), implies subordination to a verb with a different subject and is obligatorily accompanied by affixes of personal reference. In textual accounts of conversations there is a frequent alternation of the gerund and subordinator forms of ni- "to say", ni-spa and ni-pti-, which mark, respectively, the continuation of a speaker's contribution to the conversation or a transition to a contribution of his partner in the conversation. This is particularly important where there are two participants in the conversation that are both non-identical to the author or story-teller and his audience. In practice, it permits a distinction of two "third persons" and the use of lengthy fragments of text in which the subjects are not referred to lexically. Questions, exclamations and conjectures are quoted literally and are followed by ni-, which often replaces more specific verbs referring to asking, exclaiming or guessing. The concept of answering is expressed by the combination of "to say" and switch-reference. Many Quechua dialects lack a verb "to answer" and, if they do have one, its meaning is usually more specific than that of the English counterpart.

7.

Quotations in Araucanian

Other Andean languages exhibit a subdivision of the class of verbs similar to that in Quechua, assigning a special status to the verb "to say" in relation to quoted speech. In Aymara the verb sa- "to say" and its derivate sasa "saying" play a role virtually identical to that of ni- and nispa in Quechua (Hardman, Väsquez, Yapita 1988: 314). This is most certainly an effect produced by the long period of intimate contact to which the two languages have been subjected, which has led to an amazing amount of typological similarity. More remarkable is the existence of a similar situation in an unrelated language such as Araucanian (also called Mapuche or Mapudungun), which has not undergone profound influences from either Quechua or Aymara. In this section, some of the relevant Araucanian facts are outlined and exemplified. In Araucanian the verb for "to say" is pi- (or, alternatively,/e/pz'-"that-say"). Its message complement obligatorily takes the shape of a quotation, which preferably precedes the verb pi-. As in Quechua, pi- is the only verb that behaves like this syntactically, although Smeets (1989: 473) mentions three other verbal expressions that may also occur with quotation complements. Other verbs take object complements that have a nominalized verb as their nucleus. There is no equivalent of Quechua nispa. Like ni- in Quechua, Araucanian pi- covers a semantic domain that exceeds the limits of communication in the literal sense. Apart from referring to communication, it also refers to wishes and intentions, as in (17).

10

Willem F.H. Adelaar

(17) "fem-i mai", pi kifie longko; "fei be so-3sAPP say-3sone chief that

meu mai from APP

küpa-in, come-lPLURs

[kim-a-fi-in chem dengu fii mele-n] know-FUT-3o-lPLURs what matter 3p existNOM pi-ke-in APP", pi fei-chi longko. say-HAB- lPLURs then say-3s that-ADJ chief "'So it is then," one of the chiefs said. "This is why we have come. We want to know what is going on", that chief said.' In (17), pi-ke-in must be interpreted as "we want", but being a derivate of the verb pi- it is preceded by a direct quote (kim-a-fi-in chem dengu ni melen "we shall find out what is going on"). As in Quechua, Araucanian verbs contain personal reference affixes that obligatorily refer to subjects and nonsubjects. The system is more elaborate than in Quechua since it also identifies third person nonsubjects and formally distinguishes two types of pragmatic relationships between third person subjects and nonsubjects. Quotations within quotations in which all the personal reference affixes are selected according to the perspective of the corresponding subjects are frequent and may exhibit a considerable degree of complexity, as in the following example. (18) "[inche deuma I already

feyentu-fi-ii, adhere-3o-ls

fei that

meu mai, from APP

[[ [[[feyentu-pe tuchi feyentu-no-lu, fei nie-le-nge-a-i adhere-3sIMP which adhere-NEG-NOM that have-BEN-PASS-FUT3s wesha dengu]]] pi-η mele-rke-i]], pi-a-fi-mi], bad thing say-NOM exist-HRS-3s say-FUT-3o-2s pi-nge-n" pi-ke-i fei-chi werken. say-PASS-ls say-HAB-3s that-ADJ messenger 'That messenger said: "I have been told as follows: [You will tell them the following: [[I have already adhered to the cause and so then there is the (reported) saying: [[[Let them adhere! For whoever does not adhere something terrible will be kept in store.]]] ]] ]".'

The role of quotations in Andean Discourse

11

A possible free translation of (18) would be: 'The messenger said: "I was told to tell you (on behalf of my chief) that he has already adhered to the cause. And, therefore, there is said to be an agreement that you should adhere to it as well and that whoever does not will be severely punished.'"

8.

Conclusion

It follows from the above that the verb "to say" has a different status in the languages treated from that in Western European languages. Whatever its exact meaning may be, it is characteristically linked up with the use of quoted speech and thought and can very well be regarded as the sole representative of a subclass of quotation verbs. Its use and existence constitute an areal feature, since the languages at issue and the relevant expressions themselves do not visibly have a common genetic origin. In one important respect, the Andean facts are parallel to those found in Papuan languages, such as Kombai and Western Dani, where direct quotations obligatorily accompany speech act verbs (see de Vries, this volume). The necessity of preserving the form of quoted speech as it would be pronounced in a real dialogue situation and the intricate constructions consisting of quotations embedded in quotations place a heavy burden on the listener's memory and imagination. He has to keep track of the speaker's constantly changing perspective, which often moves outside the observable reality.

NOTE

The examples (3), (5) and (16) are from Soto Ruiz's beginners' course in Ayacucho Quechua; (6) and (7) are from Cuzco Quechua and were taken from the autobiography of Gregorio Condon Mamani. The examples (4) and (8) through (IS) are from the idolatry manuscript written for Francisco de Avila about 1608. It represents an extinct dialect closely akin to the former two. Wherever necessary, the orthography of the examples has been regularized in order to make their analysis more transparent. The Araucanian examples are from the memories of chief Pascual Cofia, recorded in the 20s of the present century. The original orthography has been respected as much as possible. The Quechua facts related in this article are valid for the conservative dialects of Central and Southern Peru, but not necessarily for all of the other dialects. The fragmentation into dialects does not allow us to generalize about the language as a whole. Strong influence from Spanish sometimes gives rise to a different usage, as is the case, for instance, of Condori Mamani's autobiography, where direct quotes may accompany verbs other than ni-. Even in Avila's idolatry

12

Willem F.H. Adelaar

manuscript there are such cases (for instance, quotes accompanied by rima- "to speak" instead of ni-), but these can be considered erroneous or, at the least, very uncharacteristic. The time difference between the language of the idolatry manuscript and present-day Ayacucho Quechua is not particularly relevant, because the linguistic facts at issue have remained essentially unchanged. In Ayacucho Quechua, and maybe in other dialects as well, (nominalized) object clauses can accompany the verb ni- whenever they do not express the content of a message; e.g., llapa (all) ruwa-na-n-ta (do-NOM-3ps-ACC) ni-n (say-3s) "she tells her everything she has to do" (Soto Ruiz 1987: 163). Quotes are placed in quotation marks (inverted commas), unless they are embedded in other quotes. In that case, square brackets are used with the number of brackets rising from 1 to η as one advances from outer to inner frames. Glosses used in the morpheme-for-morpheme translation of the examples have the following approximate meanings: ACC accusative case, ADJ adjectivizer, ALL allative case, APP appeal to knowledge of hearer, ASS assertion evidential, BEN benefactive case, CAUS causative, CIS cislocative, CONN phonologically conditioned connective element, CONJ conjecture evidential, DIR directional, DUR durative aspect, FUT future tense, GEN genitive case, GER gerund, HAB habitual, HESIT hesitation, HRS hearsay evidential, IMP imperative, INSTR instrumental case, INT interrogative, INCL inclusive, LOC locative case, NEG negation, NOM nominalizer, NORET no return, PASS passive, PAST past tense, PLUR plural, RESTR restrictive, SEQ sequential, SUB subordinator, TOP topic. The symbol V refers to verb. The symbols o, p, s refer to object (nonsubject), possessor and subject in combination with numbers identifying person (4 being first person inclusive); ps refers to possessor and subject simultaneously. The gloss HAB for the Araucanian affix -ke- does not seem to be appropriate in the case of the two example sentences (17) and (18). It is used here for lack of an alternative interpretation. In (18), the gloss BEN for -le- is tentative. The regular form of the affix at issue after a vowel is -(le)l- (Smeets 1989: 365).

REFERENCES Condori Mamani, G. 1977. Autobiografla. Cusco: Centra de estudios rurales andinos "Bartolomd de Las Casas". Cofia, P. 1984. Testimonio de un cacique mapuche. Santiago: Pehu6n editores. Dik, S.C. 1980. Studies in Functional Grammar. London/New York: Academic Press. Hardman, M., J. Väsquez & J.D. Yapita 1988. Aymara, compendia de estructura fonolögica y gramatical. La Paz: Instituto de lengua y cultura aymara. Smeets, I. 1989. A Mapuche Grammar. Dissertation University of Leiden. Soto Ruiz, C. 1987. Rimaykullayki, Unterrichtsmaterialien zum Quechua Ayacuchano. Republished by R. Hartmann a.o., 2d ed. Berlin: Dietrich Reimer Verlag. Taylor, G. (text ed.) 1980. Rites et traditions de Huarochiri: manuscrit quechua du ddbut du 17e siicle. Paris: fiditions L'Harmattan. Weber, DJ. 1989. A Grammar of Huallaga (Huänuco) Quechua. Berkeley/Los Angeles/London: University of California Press.

Unreportable linguistic entities in Functional Grammar

A. Machtelt Bolkestein Department of Latin University of Amsterdam

0.

Introduction

In Dik (1968: 166) it is pointed out that the maximum units to be described by a functionally oriented grammar are those linguistic entities which are in no way grammatically dependent on their preceding or following linguistic context. The size of such independent linguistic entities (ILE) may well exceed that of the sentence, as is clear from a consideration of sequences of two sentences where the second clause contains some connecting device, cr exhibits pronominalization, or forms one intonation unit together with its preceding clause, and, for example, from question answer sequences, where the answer part may not be a full sentence at all. The question about what is the upper limit of such entities is not further explored in Dik (1968). Dik insists, however, that a description along the lines of a functional grammar will not restrict itself to the sentence, and in the layout of the linguistic structure as visualized (by a tree diagram) in Dik (1968: 170f) it is indeed the case that the syntactic category sentence is not the uppermost, initial category in the hierarchical structure, but a subcategory of the initial category ILE (cf. also 1968: 191f). In this paper I will consider to what extent the expectations arising from the above claim have been fulfilled by the way in which the theory of Functional Grammar (FG) has developed since. Have studies within the FG framework

14

Α. Machtelt Bolkestein

actually investigated entities larger than sentences? What further notions have been proposed in order to describe the phenomena observed? What modifications has the theoretical framework undergone? For this purpose I will first trace the history of the notion ILE throughout Dik (1978), (1980) and (1989). I will then discuss the merits of a recent FG proposal concerning the nature of the underlying structure attributed to utterances, in the light of its potential value for describing ILE's rather than sentences. Finally I will touch upon a few of the problems which will have to be solved within any theory which adheres to the functional point of view.

1. The ILE and its replacement: extra-clausal constituents In the outline of Functional Grammar presented in Dik (1978: 15f) the claim that 'FG is meant to cover any type of linguistic expression', and that 'it is thus not restricted to the internal structure of sentences' is repeated, with reference to the earlier work. In spite of the renewal of this claim, however, the book concentrates exclusively upon the structure and properties of predications, that is, of sentences, and even, mainly, upon that of simple non-complex sentences. The notion ILE, while figuring in the index, occurs only once, in the passage just quoted. It might be concluded that FG is guilty of false pretenses in its insistence upon the importance of taking larger units than the sentence into account. However, this conclusion is not justified. One aspect to which both in Dik (1978: 127f) and in subsequent work within FG much attention is devoted, is that of the grammatical consequences of information structure, also known as Theme Rheme structure in other frameworks, as the distribution of given and new information, or as Topic Focus Articulation. In FG the latter labels are used, for two distinct 'pragmatic functions' assumed to be relevant, namely for the entity 'about' which the predication predicates something in the given setting (often old, shared information, but not necessarily), and the entity which presents the relatively most important or salient information (often new, but not necessarily) (1978: 130). Pragmatic functions are, in other words, defined as functions which specify the informational status of the constituents involved within the wider communicative setting in which they occur. This wider communicative setting can be specified as consisting of the intentions and the state of knowledge of the speaker and his beliefs concerning that of his speech partner at a particular point of the verbal interaction in a given situation. Thus, the pragmatic function of a constituent of a sentence is among other determined by its relation to other constituents in the wider context and the properties of the preceding (sometimes the following) discourse. The observation that a special pragmatic status may have consequences for the grammatical realization of the predication has inspired a number of studies since the publication of Dik (1978), and has been shown to hold true for many

Unreportable Linguistic Entities in Functional Grammar

15

languages. Correlations have been shown to exist, for example, depending on the language, with word order rules, with intonational contour, with voice variation, with the selection of special construction types such as clefting or raising, and with special morphological marking. Such studies offer ample evidence that syntactic phenomena are not autonomous, and support the claim that the grammar should be formulated in such a way so as to be able to handle units larger than the sentence. Apart from the pragmatic functions Topic and Focus, which are assigned to constituents of the predication, Dik (1978: 130f; 132f) also distinguishes a separate pragmatic function Theme, for elements which precede the predication but are not part of it: they are separated from what follows by intonation and/or pause, and may therefore be called extra-clausal. In thus differentiating two different types of 'left-dislocated' constituents, the FG approach differs from other frameworks (such as Systemic Grammar, cf. Halliday 1985) in which no principled distinction is made between intra-clausal Topic and extra-clausal Theme, or in which a one to one correlation between occupying the first position in the sentence and having a particular pragmatic status is taken for granted. The FG function Theme is not defined in terms of information structure, but in terms of relevance, namely as presenting a domain or universe of discourse with respect to which it is relevant to pronounce the following predication (Dik 1978: 138). The constituent involved need not necessarily be coreferential to any constituent within this predication (Dik 1978: 133f). The distinction between Topic and Theme is defended by Dik on the basis of several differences in behaviour which I will not go into here: suffice it to refer to a recent functionally oriented investigation of Theme constructions (Geluykens: 1988). This is a study in which it is convincingly shown on the basis of extensive attested material that in a language like English Theme constituents have a quite different function in discourse than Topics, and statistically quite frequently do not refer to shared or old information at all. Geluykens' conclusions concerning the discourse function of Themes are a refinement of, but quite compatible with the somewhat provisional definition in Dik (1978), and a confirmation of the validity of the distinction between Topic and Theme. The results of this study offer another illustration of the claim that taking discourse rather than isolated sentences into account may lead to insightful generalizations about what factors influence the use of certain syntactic constructions, and, consequently, an indirect argument in favour of the original 1968 claim. Both in Dik (1980) and in Dik (1989) the pragmatic functions Topic and Focus figure prominently in the indices and in the actual number of pages devoted to these issues, the notion of Theme somewhat less so: in (1989) it occurs only a couple of times, mainly listed alongside other types of extra-clausal constituents (: 380). Extra-clausal constituents are mentioned a few times, more or less in passing (e.g.: 45; 264f): their treatment is explicitly postponed to forthcoming work. Still, a rough indication is given of a number of functions (again called pragmatic

16

Α. Machtelt Bolkestein

functions) which they may serve. These are divided into three main functions: (i) the management of the interaction, (ii) commenting on the content of the clause proper, and (iii) the organization of the content of the expression in relation to the context in which it occurs. Most attention, however, is paid to the clause-internal pragmatic functions Topic and Focus (1989: 265-287). The account of information structure is considerably expanded and refined, and has undergone a number of changes compared to earlier work, clearly inspired by the modifications proposed in the FG inspired investigations of specific languages and specific constructions which have appeared in the meantime (e.g. De Vries 1983; Hannay 1985). Thus topicality and focality are now viewed as not necessarily mutually exclusive; and a number of subtypes both of Topic and of Focus are distinguished on the basis of the fact that these subtypes receive different marking or exhibit different behaviour in some languages (see e.g. Dik 1989:282f). Consequently, the relevance of these subtypes is language specific/ As opposed to the relevance of notions belonging to the area of information structure, the notion of independent linguistic entity has completely disappeared both from the index and from the text. Does this imply that FG has lost interest in units larger than the sentence (with the exception of matters of information structure and 'participant tracking')? This implication is contradicted by a consideration of another recent development in the theory of FG, namely the way in which it views the underlying structure of utterances (rather than that of sentences in isolation).

2.

The underlying structure of utterances: the interpersonal level

In contrast with earlier accounts of sentence structure in FG, in Hengeveld (1988; 1989; fc.) and Dik (1989: 247f; see also Reesink: this vol.), an analysis of clause structure is adopted in which various layers of structure are distinguished, which are ordered hierarchically with respect to each other. Each of these structural layers represents a different type of entity (in a formal representation, which I will not go into here, symbolized by different variables), which may roughly be described as: the sentence as an entity designating a state of affairs (the predicational level), the sentence as an entity expressing a the speaker's attitude with respect to a possible fact (the propositional level), and the sentence as an entity which is the product of a speech act, an utterance, in a certain communicative situation (the utterance or clause level).2 In other words, different levels of structure are taken to correspond to the different functions of language recognized in other approaches (e.g. Halliday 1985): thus the so-called representational or ideational function of language in Halliday is associated with

Unreportable Linguistic Entities in Functional Grammar

17

the predicational level of structure, whereas the interpersonal function is associated with the propositional level of structure and with the speech act level. As far as the hierarchical ordering of these structural layers is concerned, the highest all encompassing level, which has all other levels within its scope, is taken to be that of the speech act: the speech act level has the propositional level in its scope, and the latter contains the predicational level. Dik and Hengeveld differ mutually in what they conceive to be the highest level of the overall structure of the speech act/utterance: Hengeveld suggests that a fifth, textual, level may have to be distinguished, whereas in Dik the speech act level is the outer or upmost one. An implicit assumption of this approach is that as a rule it will be possible to assign specific linguistic phenomena to a specific level of structure according to the main function which they fulfil in utterances (although there need not be a clear one to one relation between elements of surface form and abstract elements on a particular level). This implicit assumption of 'locatability', so to speak, is the source of one of the problems to be touched upon below (2.2.). The advantages of this view of sentence structure are demonstrated by several recent and forthcoming publications, among other proposing a typology of satellites and operators (Dik 1989: 57-60; 247f); Dik e.a. fc.), and a typology of sentential complements (Hengeveld: fc.; Bolkestein 1989; fc.). I will briefly go into both typologies, and some problems connected with them. 2.1.

Satellites

Satellites are those lexically expressed constituents of sentences which are not required by the valency frame of the predicate. They may consist of single lexical items or be internally complex, for example have the internal structure of a subordinate clause.5 In Dik e.a. (fc.) it is argued that the hierarchically ordered layered structure may be used to account for differences in behaviour between at least four different types of satellites, which are attached to different layers of the layered structure. These four types of satellites are predicate satellites (σΐ), predication satellites (a2), proposition satellites (σ3) and illocutionary satellites (σ4), each attached to the corresponding structural layers respectively. Examples are the satellite constituents in (1 a-e) respectively. (1)

a. b. c. d. e.

John left on foot John left yesterday Apparently/fortunately, John has left Frankly/since you are interested, John has left Summarizing/as far as John is concerned, he left

18

Α. Machtelt Bolkestein

As mentioned above, Hengeveld (fc.) suggests that the level 4 satellites subsumed under illocutionary satellites in Dik (1989: 259) and Dik e.a., which are described as having the function of commenting on the Reason, Condition or Purpose of the speech act, might have to be further differentiated into what are called style disjuncts in Quirk e.a., that is into illocutionary satellites (σ4) on the one hand, and a fifth type of satellites, clause satellites (σ5) on the other. The former have the function of specifying the manner of the speech act, such as frankly, honestly, seriously, and the like, and the latter have the function of specifying the felicity of the utterance in its discourse context, such as for example in brief, summarizing, finally (cf. (l)e). If we regard this description of the function of clause satellites as another way of saying that their function is to manage the interaction process, Theme constituents such as as far as...is concerned also qualify for σ5 status. This difference in approach is connected with the different view which Dik and Hengeveld take of the status of illocution in the underlying structure as a whole: in the account of the former it is considered to be a speech act (utterance?) operator, which determines the type of speech act, event or product on the very higest layer; for Hengeveld basic illocution has the status of an abstract predicate frame designating the specific semantic and/or pragmatic relation between the speech participants and some content; in the latter view there is an empty position on the highest level of the underlying structure for utterance operators and satellites. For the present purpose I will not go into this problem further. Without repeating the criteria which justify this distinction between these (at least) four types of satellites (which are partly already given in earlier literature on differences in behaviour between adjunct and disjunct adverbials), I want to point out that Dik e.a. stress the fact that the two 'lowest' level satellites are clause internal, whereas the higher types, from σ3 up are extra-clausal constituents. Their extra-clausal status is confirmed by phenomena like intonational contour, by the fact that Topic Focus distribution takes place within the predication proper and does not include such constituents; and by the fact that operators of the lower levels, such as tense, sentence negation and inferential modality, also have no scope over them: the other way round, such lower level operators fall within the scope of these higher satellites, as may be seen from (2): (2)

a. b.

Frankly, fortunately, John had not left Since you ask, John might be ill

As observed above, basic sentence type distinctions such as the distinction between Declarative, Imperative and Interrogative sentences are in FG (see Dik 1989: 254f) described as basic illocutions (ILL). They are taken to have the status of either operators on the highest, speech act, level, or alternatively of abstract predicate frames (Hengeveld 1989; fc.). Illocutionary satellites fall outside the scope of the basic illocutionary force of sentences, as shown by (3):

Unreportable Linguistic Entities in Functional Grammar (3)

19

Frankly, why did John leave?

The fact that such satellites are outside the basic illocution is problematic in view of the tacit assumption in FG that operators of a certain layer have satellites of that same layer within their scope. The fact that ILL does not have scope over illocutionary satellites such as frankly seems an argument in favour of attributing to ILL a status on a lower layer of the hierarchical structure than that accorded to it in Dik (1989: 254f). This is supported by a consideration of complement constructions, another area where the hierarchically ordered layered structure account seems to be a fruitful approach. 2.2.

Reportable and unreportable speech acts

Both Hengeveld (1989a: 146; 156; fc.) and Dik (1989: 248f) argue that given the distinct structural layers and types of entities represented by them, sentential complements may also be subcategorized according to the type of entity they represent. Matrix predicates of different semantic subclasses govern different types of entities, and this has to be specified in their lexical entry, since it may have consequences for further properties of these complements. Thus verbs of direct perception such as to watch will govern either first order entities (concrete objects) or second order entities (events), verbs of emotion such as to regret may govern either events or third order entities (propositions), in the latter case carrying a meaning like 'to think regretfully that...'; whereas verbs of speech will govern either utterances (when governing direct speech) or propositions (when governing indirect speech). In Bolkestein (1989; fc.) some of the implications of this claim are explored on the basis of investigation of complement clauses in classical Latin. It is concluded that a number of behavioural criteria support a differentiation among complement types along the lines of the layers which they represent. The clearest of these is the distribution of level 3 and level 4 satellites (see 2.1.) over the various complement types: such higher level satellites only occur in complements of verbs of speech and thought, but are absent from complements of verbs of emotion, of direct perception and of happening: (4)

a. b.

he said that, since she wanted to know, he had left (σ4). he said that John was probably leaving (σ3)

(5)

a. b.

?he regretted that, since she wanted to know, he had left ?he watched John probably leave

In view of the possibility for indirect speech complements of verbs of speech to contain extra-clausal constituents presumably belonging to the highest level, it is

20

Α. Machtelt Bolkestein

not correct to state that such complements represent proposition entities. In fact, complements of such matrix verbs may themselves be classified according to the basic illocution (Declarative, Imperative, Interrogative) which they carry: this classification is useful in view of the fact that in many languages, among which Latin, certain formal and behavioural properties correlate with these distinctions, such as (i) the choice of subordinating conjunction, (ii) the existence of semantic restrictions for Imperative complements as to the controllability of the state of affairs designated, and (iii) as to the presence of truth value and evaluative satellites (Bolkestein 1976; fc.; for the notion of controllability see Dik 1989: 96). This means that complements of verbs of speech should be analyzed as propositions plus an illocution rather than as bare propositions, and, given the compatibility with some types of extra-clausal elements pointed out above, even more than that. In certain cases it can be argued that not only basic illocution, but even non-basic or 'converted' illocution (for the notion of illocution conversion see Dik (1989: 257f)) is relevant for formal properties of the embedded clause: thus, in classical Latin, interrogative complements which have the illocutionary force of statements (so-called rhetorical questions) are realized by a different construction than real questions for information are: the verb of the former will be realized as infinitive, that of the second as a finite verb in the subjunctive mood, in spite of the fact that in direct speech they are formally not distinguished. However, as observed by Coulmas (1986: 9), indirect speech complements cannot contain all material which may be present in direct speech utterances. For example, utterances with a fourth basic illocution, namely that of Exclamation (which in some, but not all languages forms a separate sentence type), cannot be reported by indirect speech complement at all: (6)

a. b.

What cheek she has! *He exclaimed that what cheek she had

Some other sequences of more than one clause which in view of their intonation form a linguistic entity larger than a single sentence cannot litterally be reported in indirect form either. This is the case, for example, with warnings and threats (and also with contingent promises and predictions) of which the first clause has the verb in the imperative mood (but which cannot be described automatically as having the basic illocutionary force of an Imperative). The form in which such ILE's have to be reported has to be turned in that of an ordinary Declarative conditional construction: (7)

a. b. c.

drink that and I'll kill you *he threatened that she should drink that and he would kill her he threatened that if she drank that, he would kill her

Unreportable Linguistic Entities in Functional Grammar (8)

a. b. c.

21

drink that and you'll die ?he warned her that she should drink that and she would die he warned her that if she drank that, she would die

Furthermore, in spite of the possibility to contain certain extra-clausal constituents such as those demonstrated in (3) and (4), the presence of various types of extraclausal constituents such as those having the function of Initiator (well, now, see Dik (1989: 264f)) or other starters (like adversative however, but then,) or the sentence-final particle though), vocatives having the function of Address {Ladies and Gentlemen) and discourse markers such as oh, allright, (ah) yes, and Tags which have the function of modifying a Declarative illocution into Interrogative one (isn't it) leads to unwellformed results. Dutch examples are nee, nou, eh 'no, well, eh', (ja) maar '(yes) but', etc. Many other elements which parenthetically refer to either the speaker's or the hearer's attitude or state of knowledge (look, you know, I mean, I see, but also evaluative comments like in Dutch fijn 'great'or jammer 'a pity') are excluded as well. If such extra-clausal elements are present, the effect will be strange or change the degree of indirectness of the reported speech into the type of free or semi-indirect speech which would be better without the subordinating conjunction, cf. (9): (9)

a. b. c.

?he said that well, he was hungry ?he said that ladies and gentlemen, this was a festive occasion ?he said that it was a festive occasion, wasn't it

This shows that there is a difference, perhaps in the degree of extra-clausality, between Theme constituents and speech act modifying satellites such as since you...., on the one hand, and on the other, certain other elements which seem to be more interaction oriented and more characteristic of dialogue than of monologue discourse. Within the hierarchical utterance structure such differences between extra-clausal elements could only be accounted for if further distinctions are made on the highest structural level. Moreover, the extra-clausal status of some elements excluded from occurring in indirect discourse is doubtful, if certain criteria (e.g. the criterion of intonational contour) are maintained as decisive. A case in point is offered by Latin leading question particles (as opposed to the neutral clitic particle -ne) such as num, which indicates that the speaker himself expects or wishes to have the question answered in the negative, and the particle nonne which indicates the expectation or wish for a positive answer. In indirect speech questions governed by matrix verbs of asking such as rogare 'to ask', such particles do not occur with the same function. The same holds for the Dutch expressions toch (zeker) niet (a complex particle combination suggesting that a negative answer is plausible) and toch (zeker) wel (positive answer), which functionally resemble extra-clausal tags in English, but are

22

Α. Machtelt Bolkestein

neither sentence initial nor extra-clausal from the intonational contour point of view: (10) a.

hij is toch (zeker) niet weggegaan? he aux complex Q-particle gone-away 'he hasn't left, has-he?'

b. ?ze vroeg of hij toch (zeker) niet was weggegaan she asked whether he complex Q-particle aux gone-away 'she asked whether he hadn't left, had he' (11) a.

b.

Nonne abiit? Q-particle leave:PF/3P 'he has left, hasn't he?' ?rogabat nonne abiisset ask:IMPF/3P Q-particle leave:PF/SUBJ/3P 'she asked whether he had left, hadn't he'

Such particles express the speaker's attitude with respect to the possible truth value of the proposition questioned. They could, therefore, conceivably be argued to function on the propositional level rather than on the speech act level, as opposed to particles and other expressions of which the function is more clearly that of managing the interaction, such as for example the Latin sentence initial particle cm ('is the conclusion justified that...', 'does that mean that...'; in Dutch the particle dan 'then' may occur with a similar function in questions), which introduces questions which have the function of checking whether an unexpected potential implication of a preceding utterance in the discourse is indeed justified: (12) SI: te dum vivebas noveram you while live:IMPF/2P know:PLPF/lP Ί knew you while you were alive' S2: An me mortuum arbitraris? Q-particle me dead think:PR/2P 'do you believe I am dead, then?' (from a comedy by Plautus, True. 164) The particle an illustrated in the second clause in (12) could be argued to be more context oriented and more interaction oriented than num or nonne (see 11): the utterance introduced by it always has a close relation to some preceding utterance or act, usually (but not necessarily) performed by another speech participant. Like num and nonne, it also is excluded from being embedded with this same function

Unreportable Linguistic Entities in Functional Grammar

23

in Interrogative complements. In the case of an, this could perhaps be accounted for by its characteristic dialogic function. However, none of the question particles are extra-clausal, and the same holds for many other particles which typically have the function of explicitating the relation between the clause which contains them and some preceding or following clause. It is often difficult to decide whether such elements (I call them particles, as an indiscriminate waste basket label) relate text units (e.g. paragraphs) or rather single clauses, and in the latter case whether they relate these clauses on the speech act level or on a lower, representational ('semantic' or 'logical'), level. However, it is still possible to show that some of them are relatively more interaction oriented than others, as has been argued in Kroon 1989 for the Latin pairs enim vs. nam, and ergo vs. igitur, usually called causal and consecutive particles respectively. Kroon shows convincingly that the members of these pairs are wrongly treated as more or less interchangeable in the handbooks. However, she also shows that this difference in interactionorientedness is relative rather than absolute: one of the properties of such sentence connecting devices is that they function in more than one way at the same time, as is often observed in particle research. Even if this were not the case, with respect to the description of such interaction managing particles within the layered structure account, we are confronted with the 'locatability' problem alluded to in 2.1: in view of their function we would perhaps want to situate them in the underlying structure ascribed to utterances on the outermost layer, which also hosts other, extra-clausal, discourse markers. At the same time, since they cannot be analyzed as extra-clausal themselves, the criteria for recognizing satellites belonging to these levels do not apply to them. The question remains both as to what status should be assigned to them, and whether, whatever they are, they should be treated as situated on any specific layer of structure. Within FG there are as yet no explicit proposals as to how such elements should be described in agreement with the sentence model as a whole. Various options are conceivable: (i) as abstract predicates (on some level), (ii) as operators characterizing the clause which contains them as a specific type of entity (an option which neglects their basic relating function), or (iii) more or less parallel to the way subordinating conjunctions are treated, as the product of expression rules which have a number of primitive functional relations between clauses or entities of another nature (whether or not with the requirement that these entities should belong to the same layer of structure) as their input. The lack of attention for such relations within texts and the lack of explicitness of the theoretical framework with respect to these alternatives means that there are quite a number of barriers to take in order for the theoretical framework to account for entities larger than sentences.

24 3.

Α. Machtelt Bolkestein Concluding remarks

Has FG actually looked at entities larger than sentences which could be called ILE's in the original 1968 sense? Yes and No. There are (at least) two ways in which sentences may exhibit linguistic properties which are sensitive to (and therefore not independent of) the surrounding discourse, that is, may have linguistic properties which are due to the fact that they are part of a larger whole: (i) by the fact that they exhibit a certain information structure which may have consequences for certain formal phenomena, and (ii) by the fact that they fulfil a certain function with respect to the main goals of the speaker and to their surrounding discourse. Concerning the first type of cohesion phenomena the discussion within FG is in full swing, and has already resulted in a further elaboration of Topic Focus articulation. With respect to the second type of cohesion, FG is still at the start, and may profit considerably from proposals within neighbouring disciplines such as text grammar, discourse analysis or rhetorical structure theory. At the same time, in spite of many unsolved problems, the layered structure account of utterances recently developed in FG makes possible a consistent and formal treatment of the properties of intra- and extra-clausal constituents, and seems to be able to handle at least some properties of utterances which are connected to their discourse function. The recognition of the importance of the speech act level (which constitutes a third area in which linguistic properties of sentences may not be independent from factors outside their own domain) has already opened the way for further developing a description of phenomena connected with illocution (for attempts in that direction see e.g. Vet fc..; Risselada 1989). For FG to make come true its final aim, it is to be hoped that the trend will extend itself to the study of the functional relations between sentences within context.

Unreportable Linguistic Entities in Functional Grammar

25

NOTES 1. In a recent WPFG (Working Papers in Functional Grammar) Mackenzie & Keizer (1990) and Hannay (1990) discuss the potential (ir)relevance of the notion Topic for a description of word order in English. 2. I would speak of the product of a speech act rather than of the speech event or the speech act. In Hengeveld (1989; fc.) the highest level variable is said to represent the speech event rather than a product of speech activity. This seems to me unfortunate: the structure of the speech event is only relevant in as far as it is reflected in the structure of the utterance or stretch of text (sequence of utterances) produced. Dik (1989: 255f) describes the variable involved as a variable for the speech act, but also speaks of expressions, utterances and clauses (which are products rather than acts or events). Although I will continue to use the term 'speech act level' this should, to my opinion, be interpreted as 'the level of the speech product.' 3. Although it may not be unimportant, I will not go into the fact that there is no consensus within FG concerning the formal representation of satellites: whereas Hengeveld, following Vet fc., represents them as restrictors on predications (propositions etc.) in a way comparable to restrictors on term level, Dik represents them as fulfilling a Semantic Function within the predication (proposition etc.) as a whole.

REFERENCES Bolkestein, A.M. 1976. The relation between form and meaning of Latin subordinate clauses governed by verba dicendi. Mnemosyne 29, 155-175; 268-300. Bolkestein, A.M. 1989. Latin sentential complements from a Functional Grammar perspective. Cahiers de l'Institut de Linguistique de Louvain 15. 1-4, 41-52. Bolkestein, A.M. fc. Sentential complements in Functional Grammar. In: J. Nuyts, A.M. Bolkestein & C. Vet (eds.) Layers and Levels of Representation in Language Theory. Amsterdam/ Philadelphia: John Benjamins, p. 71-100. Coulmas, F. 1986. Reported Speech: some general issues.In: F. Coulmas (ed.) 1986, Direct and Indirect Speech. Berlin/New York/Amsterdam: Mouton De Gruyter, p. 1-28. Dik, S.C. 1968. Coordination. Amsterdam: North-Holland Dik, S.C. 1978. Functional Grammar. Amsterdam: North-Holland (3d printing 1981 Dordrecht: Foris). Dik, S.C. 1980. Studies in Functional Grammar. London/New York: Academic Press. Dik, S.C. 1989. The Theory of Functional Grammar. Part I: The Structure of the Clause. Dordrecht: Foris. Dik, S.C., P.C. Hengeveld, E. Vester and C. Vet fc. The hierachical structure of the clause and the typology of adverbial satellites. In: J. Nuyts, A.M. Bolkestein & C. Vet (eds.) fc. Layers and Levels of Representation in Language Theory. Amsterdam/Philadelphia: John Benjamins, p. 2570. Geluykens, R. 1988. Left-dislocation in English Discourse. Ph.D. Diss. University of Antwerp. Halliday, MA.K. 1985. An introduction to functional grammar. Baltimore: E. Arnold. Hannay, M. 1990. Pragmatic function assignment and word order variation in a Functional Grammar of English. Working Papers in Functional Grammar (WPFG) 38.

26

Α. Machtelt Bolkestein

Hengeveld, P.C. 1988. Illocution, mood and modality in a functional grammar of Spanish. Journal of Semantics 6.3/4, 227-269. Hengeveld, P.C. 1989. Layers and operators. Journal of Linguistics 25.1, 127-157. Hengeveld, P.C. fc. The hierarchical structure of utterances. In: J. Nuyts, A.M. Bolkestein & C. Vet (eds.) fc. Layers and levels of Representation in Language Theory. Amsterdam/Philadelphia: John Benjamins, p. 1-24. Kroon, C. 1989. Causal connectors in Latin: the discourse function of nam, enim, igitur and ergo. Cahiers de l'Institut de Linguistique de Louvain (CILL) 15, 1-4. Mackenzie, J.L. & M.E. Keizer 1990. On assigning pragmatic functions in English. Working Papers in Functional Grammar (WPFG) 38. Risselada, R. 1990. Illocutionary function and functional illocution. Working Papers in Functional Grammar 34. Vet, C. fc. Asymmetries in the use of tense and modality. In: J. Nuyts, A. M. Bolkestein & C. Vet (eds.) Layers and Levels of Representation in Language Theory. Amsterdam/Philadelphia: John Benjamins, p. 123-138.

Issues in Functional Discourse Analysis

Teun A. van Dijk Program of Discourse Studies University of Amsterdam

The highest aim of a functional grammar of a particular language is to give a complete and adequate account of the grammatical organization of connected discourse in that language. (Simon C. Dik, The Theory of Functional Grammar, Part I, 1989, p. 12).

0.

Introduction

In a paper honoring Simon C. Dik and his seminal work on Functional Grammar (FG) it may be expected that special attention be paid to the functional properties of language and language use. Similarly, given my own framework of research, it may also be expected that such a functional approach should focus on discourse. I shall fulfill these expectations by examining some of the implications of what may be called 'Functional Discourse Analysis' (FDA). FDA has a linguistic component, which deals with grammatical and other functional relations of textual structures or strategies, and a broader, interdisciplinary component, which analyzes the functional relations between these textual structures and various structures of the 'context', such as those of cognition, interaction, and even the structures of the societal or cultural macro-level. The first component will be called 'Functional Text Analysis' and the second 'Functional Text-Context Analysis'. 'Functional Discourse Analysis'

28

Teun A. van Dijk

is used to denote the integration of both approaches. Following common practice, I shall use the term 'functional' to denote both kinds of relationship, although obviously different kinds of 'function' are involved. Functional Text Analysis and Functional Grammar not only overlap, they also have common roots and share links with related approaches, e.g., in Tagmemics (Pike 1967; Grimes 1975; Longacre 1977), and especially in Czech structuralist work on Functional Sentence Perspective (FSP), for instance on such important but still elusive notions as Topic/Theme, Comment/Rheme and Focus (DaneS 1974). The Czech linguists were among the first to analyze systematic relations between sentences of texts, and emphasized the dependency of sentential topics from typically textual phenomena such as the dynamic relationships between 'old' and 'new' information in sentences and sentence sequences. It is this discussion that also inspired much later work, both in Functional Grammar (Dik 1978, 1980, 1983, 1989; Bolkestein, de Groot & Mackenzie 1985), as well as in other (typological, systemic) approaches to discourse grammar, and related work at the boundaries of linguistics and cognitive psychology. (Givön 1983, 1989a, 1989b; Haiman & Thompson 1988; Halliday 1985; Steiner & Veltman 1988; Tomlin 1987a). Thus, it is one of the hallmarks of functional approaches to language to emphasize the discursive nature of language use and to analyze (at least some) sentential structures as being functionally dependent on the structures of text and talk. Unfortunately, among many grammarians, especially in the Netherlands, this discourse orientation is still often neglected for a more restricted focus on the structures of sentences: In everyday practice, sentence grammar continues to be seen as the proper task of the linguist, whereas the study of discourse is relegated to the domain of, e.g., conversation or argumentation analysis. Indeed, 'discourse grammar' and 'text linguistics' may now have become salonfähig in linguistics, or even popular in some countries, they are still far from being mainstream. It cannot be the task of this paper to review the vast literature in functional linguistics that does pay explicit attention to discourse structures. Rather, I shall select a number of specific functional properties of texts that have received less (or, in my opinion, less adequate) attention in the literature, and show that an adequate theory of functional relations also needs a text-context component. Many important issues in the functional study of discourse, such as the analysis of topic, comment, foregrounding, focus, and similar notions extensively discussed in the literature, will be ignored here. Indeed, whereas these linguistic approaches may yield satisfactory descriptions of such functional relations within sentences and texts, a text-context analysis of functional relations also allows us

Issues in Functional Discourse Analysis

29

to provide explanations of such textual structures, for instance in terms of the details of their cognitive processing or their social functionality.

1. Functional Text Analysis One major task for a functional analysis of text structure is to systematically describe the functional relationships between sentences of texts. These relationships may theoretically be established for each grammatical or other level of analysis, that is, in terms of a classical phonological, syntactic, or semantic description, but also in the framework of a pragmatic, stylistic, rhetorical or superstructural (e.g., argumentative, narrative, etc.) account of text structure. Although several of these levels or 'dimensions' of discourse structure are intimately related, I shall focus first on semantic relationships. 1.1.

Functional text semantics

If we assume that the semantics of texts should (also) be formulated in terms of propositions, relations between sentences are to be defined in terms of relations between propositions (van Dijk 1977). These relations are of several kinds and underlie the definition of one major property of textual meaningfulness, viz., (local) coherence. First, we may have relations between concepts of subsequent propositions, for instance those of conceptual identity (synonymy) or inclusion (hyponymy, hyperonymy). Although such relations often obtain in texts, they are neither sufficient nor necessary to define coherence. To define the structural semantic notion of textual coherence, propositions should be related as a whole, and not only in terms of their predicates or arguments. The primary conditions of textual coherence are referential or extensional. That is, discourses are coherent first of all if their respective sentences refer to 'facts' (states of affairs, events, etc.) that are related in some situation (possible world, model, etc.), for instance by a relation of time, conditionality or causality, as in the trivial example (1)

John is sick. He can't come.

Here the first sentence denotes a fact that is a cause of the fact denoted by the second sentence. Similarly, the fact denoted by a previous sentence may 'enable' the fact denoted by the second sentence: (2)

John went to the movies. He didn't like the film.

30

Teun A. van Dijk

(3a) John went to the swimming pool. He didn't like the water. (3b) ?John went to the movies. He didn't like the water. Obviously, these coherence relations are defined relative to the knowledge of the world of the speaker (or hearer), e.g., as represented in scripts of culturally variable episodes (see below). In other words, coherence in this case is not directly defined in terms of propositional relations, but indirectly through (known) relationships in the world. Or even more succinctly: A text is coherent if it describes facts that are known or believed to be related. In more formal or cognitive terms, thus, a text is coherent if it can be interpreted in a (formal, mental) model. The notion of a mental model is crucial for a functional analysis of discourse, and will be discussed in more detail below. One of the well-known properties of local semantic coherence, viz., co-reference, is dependent on such a relation between facts in a model, and obtains when elements (individuals, properties, relations) of such related facts are identical. It should be repeated, however, that coreferential identity, or its functional correlate, viz., topic continuity, is neither a sufficient nor a necessary condition of local coherence: We may tell an incoherent story about the same person, and a coherent one about many different people. Besides this type of referential coherence, sentences in texts may also be functionally coherent. In this case, the condition is not extensional but intensional: the relation may be conceptually defined in terms of the propositions themselves: (4)

John can't come. He is sick.

In this case, the first sentence doesn't describe a fact that conditions the fact described by the second sentence, as was the case in (1). Indeed, the order of the sentences seems to run against the 'natural order' of the events: A consequence is mentioned before its cause or reason. If natural order representations are the unmarked, 'normal' way of describing events, a different order is the marked form and usually has a specific function, as is also the case here. I shall come back to this notion of 'natural order' below. The second sentence not only describes the cause or reason of a fact mentioned earlier, but by doing so it is assigned the function of an Explanation. Explanations, thus, are a functional category of textual structure, because they are defined in terms of their relations to other propositions in the text. In the

Issues in Functional Discourse Analysis

31

same way, next sentences (or rather, propositions) may thus 'specify', 'generalize', 'contrast with', 'paraphrase', 'contradict', 'conclude from' or 'give an example of earlier propositions in the text (for an early approach to such functional or 'rhetorical' relations, see e.g., Grimes 1975). In other words, functional relations tell us something about the information structure of a text, and not about the world the text is about. Whereas such a functional analysis of coherence is intuitively satisfactory, it is rather difficult to make explicit in formal terms. Whereas referential coherence can be defined in the relatively precise terms of fact relations in formal or cognitive models, functional relations are more elusive. It is not surprising, therefore, that the most advanced current work on such textual structures still has theoretical shortcomings, e.g., because it confuses referential and functional relations of coherence, as well as local and global functional relations and categories (Mann & Thompson 1988; Mann, Matthiessen & Thompson 1989). Some functional relations allow formal definition. For instance, Specification, Generalization and (in part) also Conclusion may be defined in terms of semantic entailment or its cognitive counterpart, e.g., Β is a Specification of A if Β (semantically, conceptually, cognitively) entails A, as in the pair 'John is sick. He has the flu'. Similarly, Paraphrase or Contradiction may be (formally) defined in terms of mutual entailment and negation. Note that these intensional definitions of functional relations also have their extensional correlates: Also the facts denoted are related (the flu is an illness, and if John has the flu he is also sick). Explanations, however, are more complicated, if only because they seem to be dependent on a specific referential relation, viz., that of cause and consequence. Thus, explanations in general are explanations of facts, not explanations of previous information: The second sentence/proposition in (4) does not, stricto sensu, explain the first sentence/proposition. However, as a semantic discourse function, the functional relation of an Explanation consists in adding further information of a specific kind, viz., about the causes or reasons of a previously mentioned (and hence known) fact. This approximate definition is consistent with the most general functional relation of text semantics, viz., Addition. However, Explanation remains rather problematic as a semantic function of discourse, because unlike Generalization or Contrast the definition is apparently not merely given in terms of intensional propositional relations. These functional relationships are proper discourse functions for independent sentences. Thus, they cannot usually be reduced to clausal relations within compound sentences. We cannot say, for instance, John is sick, and he has the

32

Teun A. van Dijk

flu, whereas the sentence John doesn't come because he is sick is a discourse Explanation only if the information 'John doesn't come' is presupposed (and repeated in the present sentence), e.g., because of earlier information in the discourse. Whereas referential coherence links are fine within a compound sentence like John is sick, and (so) he can't come, we cannot say: John can't come and he is sick if we want to convey that his sickness is the reason for his absence. Note that these functional relations not only hold across sentence boundaries, but also across turn boundaries in conversation, for instance in order to explain or specify earlier information (e.g., for other speech participants): (5)

A: John can't come. B: He is sick.

(6)

A: John is sick. B: He has the flu.

Functional semantic relations in discourse not only hold between single, subsequent sentences, but may also obtain between a proposition and a set or sequence of propositions. In this way, a proposition may function as a Conclusion, Summary, Example or Contradiction of a whole stretch of discourse. Similarly, they need not be limited to the level of local coherence relations among the propositions expressed in the subsequent sentences of a text, but may also characterize relations between inferred macropropositions 'summarizing' a text fragment. Thus, the sentences that follow a summary, for instance the headline or lead of a news report in the press, may individually or jointly be assigned to the functional category of a Specification (van Dijk 1988). In other words, functional relations not only characterize the propositional microstructure of the text, but also its overall semantic macrostructure. If such functional relations between macropropositions become conventionalized, we analyze them in terms of superstructures, that is, as text schemata that define the overall form of a discourse genre, such as narratives, news reports, conversations, or scholarly papers, which feature such well-known functional categories as Summary, Introduction, Setting, Complication, Discussion, Evaluation or Conclusion (van Dijk 1980). Depending on the theoretical framework, semantic functions may be analyzed in a 'pragmatic' way if we see these relations not as semantic relations between

Issues in Functional Discourse Analysis

33

propositions or as semantic categories, but as acts, that is, as something people do: Generalizing, Specifying, Explaining, Contradicting, Paraphrasing or Concluding (van Dijk, 1981). However, note that these acts are not the usual illocutionary acts like assertions and promises, each with their own appropriateness conditions, but at most 'propositional acts'. Hence, they belong to a semantic account, although such an account may be coupled with a pragmatic or interactional one, viz., as a (mental) 'act' that must be accomplished in order to accomplish an illocutionary act. Although mental 'acts' are not acts but processes, it is not surprising that functional categories do seem to have act-character, because they precisely specify the function of a proposition or speech act, and such functions are often intuitively described in terms of verbal acts: (7)

Ί have no time to go to that conference', she added (explained, specified, etc.).

Here, an assertion is categorized by its function in the text, which seems to suggest that a specific sub-genre of the speech act of an assertion is being accomplished. This is however not the case in a proper pragmatic analysis of speech act functions (for a detailed discussion and classification of such meta-communicative acts, see Kreckel, 1981). 1.2.

Functional pragmatics

We see that much in the same way as sentences or propositions may be functionally related, we also find functional relations between speech acts in speech act sequences. Indeed, next speech acts may also specify, generalize, correct, contradict or explain previous speech acts: (8)

Could you do something for me? Could you please help me change my tyre?

(9)

Could you bring me some coffee? No, bring me some tea, please!

Whereas I have doubts about the functional nature of semantic Explanation, pragmatic Explanation is a functional explanation in the strict sense: a speech act may indeed be intended as an explanation of a previous one, as in (10) Can you tell me the time, please? I forgot my watch.

34

Teun A. van Dijk

Such an explanation, thus, pertains not to the reasons or conditions of a fact referred to by the previous sentence or speech act (telling the time), but to the appropriateness conditions of the previously accomplished speech act itself: The previous request is appropriate if the speaker has no obvious other way to realize the goal of the request. Relations between actions, and hence between speech acts, can be assigned to a number of specific additional functional categories, however. Thus, speakers may challenge previous speech acts, usually by questioning their appropriateness conditions or their (pragmatic) implications or implicatures: (11) A: Can you please tell me the time? B: But, you have a watch yourself! (12) A: Congratulations! B: Why? Even more than at the semantic level, functional analysis of speech act sequences is tricky. That is, we should not confuse speech acts with speech act functions! The pragmatic function of a Question (or of Questioning) is not the same as asking a question, but a functional category denoting the class of all speech acts that may be used to question a previous speech act, including not only questions but also assertions or accusations. Sometimes, as in (12), the speech act of a question may also have a Question function relative to the previous speech act. This is not the case, however, in the following example, where a subsequent question does relate to the previous speech act, but it does not question the appropriateness of the previous speech act as such: (13) A: The money is gone! B: Are you accusing me? In this case, the function of the speech act of Β is to ask for the pragmatic implication of the previous speech act, viz., whether the assertion is intended as an indirect accusation. Such a function may be called Ascertaining. Speakers may also have their own version of making sure that their previous speech act is understood correctly: (14)

You get a new computer. That's a promise.

The second speech act may in this case function as a Qualification of the previous speech act. Interestingly, together the two speech acts function as a (macro) speech act, viz., as a promise. Although the second speech act may

Issues in Functional Discourse Analysis

35

itself intuitively also be regarded as a promise, there are formal reasons not to analyze it as such: unlike promising speech acts it does not refer to a future action of the speaker, and the use of that's excludes it as a self-referential speech act (a performative). Also at the pragmatic level, functional relations may obtain between speech acts and sequences of speech acts or macro-speech acts: One speech act may question, denounce, or correct a whole sequence of previous speech acts, together functioning as an accusation or a threat, for instance. 1.3.

Rhetorical functions

Several authors have dealt with functional relations in discourse in terms of 'rhetorical' structures (Grimes 1975; Meyer 1975; Mann & Thompson 1988). I find this use of 'rhetorical' confusing and prefer to reserve it for the specific structures specified by a (partial) theory of rhetoric, such as traditional 'figurae' at several levels of linguistic description, e.g., alliteration, rhyme, parallelism, metaphor or irony. However, some semantic and pragmatic functions also seem to require rhetorical analysis in the strict sense. For instance, Contrast may be defined in semantic terms, but also may have a rhetorical function if it is used to enhance the effectiveness of the discourse. Similarly, exaggerations or mitigations may also have both a semantic and a rhetorical function, as is the case in the following sequence, where the second sentence expresses a weaker claim than the first, a mitigation that also has a rhetorical function in a persuasive context (see below): (15)

He is a bloody racist. At least he doesn't like to have a Black boss.

One of the crucial differences is that semantic functions necessarily require a definition in terms of propositional relations, as in example (15). In a rhetorical analysis, however, this is not the case. Exaggerations, mitigations, or litotes are defined with respect to what is actually meant or expected in a specific situation, and therefore suggest a form of substitution, which is one of the major meta-operations of rhetorical figurae. Hence, in rhetorical operations we do not necessarily have a relation between actually expressed propositions, but between a proposition and a potential (expected, intended) proposition. Obviously, however, semantic Mitigation may also have rhetorical functions. Both may play a fundamental interactional role in face keeping or self-presentation strategies, as is also the case for well-known semantic moves such as Apparent Denials or Apparent Concessions, e.g., typically in racist talk (van Dijk 1984, 1987a):

36

Teun A. van Dijk

(16) I have nothing against Blacks, but I don't like them as my boss. (17) There are also smart Blacks, but I still don't like them as my boss. These examples are interesting also because they show other aspects of functionality in discourse. Sentences (16) and (17) do not as such have a semantic function in isolation, nor are Denials and Concessions speech acts (no more than Explanations or Generalizations). Rather, they may have a functional role, as a move, within an overall strategy, for instance in a strategy that combines negative other-presentation of minorities with positive self-presentation as a tolerant citizen. Semantically speaking, the relations between the clauses are merely an example of a contrast, as is also expressed by the typical but at the beginning of the second clause. In some cases, several functions operate at the same time, as for instance in the following example taken from a racist interview, where we find both a semantic, a rhetorical and a strategic (interactional) Contrast: (18)

We always have to work hard for our money. They simply get their money from welfare!

We see that in actual discourse analysis, several levels and dimensions of description may exhibit similar functions. It is however important for theoretical reasons to try to make careful distinctions between these different functions. Thus, a semantic function of Contrast may be used as an effective rhetorical figure, and such a figure may again function as a move within an overall strategy of self-presentation. It is one of the tasks of future work on functional relations in discourse to make these different types of functions more explicit for each level of discourse analysis.

2.

Cognitive functions

A broader functional discourse analysis also examines the functions texts and their structures have relative to their 'contexts'. Since meanings, and hence semantic functions, have a cognitive basis, we also need to explore what cognitive functions may be involved in discourse. Also classical analyses of Topic-Comment relations often (intuitively) operate with cognitive notions, such as old and new information or knowledge. Similarly, definitions of notions such as presupposition and implication also have both a more abstract semantic and a more cognitively flavored analysis. In such a cognitive analysis we may, rather trivially, 'translate' the semantic analysis given above of such functions as Generalization, Specification, or

Issues in Functional Discourse Analysis

37

Contrast, into more cognitivistic terms, viz., by identifying propositions not as abstract objects, but as units of memory representations of a text (van Dijk & Kintsch 1983). In that case, Generalization could be redefined as a cognitive operation on knowledge structures, whereas a Contrast, which is merely implicitly 'expressed' as a relation in a semantic structure of the text, may even be explicitly represented as such in memory, viz., as an actually intended contrast between two propositions, that is, as a proposition in its own right (taking two other propositions as its arguments). 2.1.

The cognitive basis of semantic FG categories

Instead of further exploring these rather obvious links between abstract semantic and more or less empirical cognitive properties of meaning, it is more interesting for the discussion in this paper to have a closer look at other functional properties of a cognitive approach to discourse. One interesting link between a cognitive study of discourse processing and a functional study of language is not limited to discourse analysis but specifically also pertains to the foundations of both Functional Grammar and functional discourse theories. Thus, FG and other functional grammars embody (a variant of) classical 'case' analysis (Fillmore 1968), and distinguish such semantic role categories as Agent, Patient, Experiencer, Object, etc. (Dik 1978, 1980, 1989; Givön 1979). Strictly speaking, such a case/role analysis cannot be properly grounded in an abstract meaning analysis because an Agent is not an 'agent' of a proposition, but of the action denoted by the proposition; the same is true for other semantic roles. However, in linguistics and cognitive theory we do not deal with 'real' referents in the world, that is, with actions or events, but with their cognitive counterparts, viz., with the ways such events, actions or episodes are represented by language users. Thus, words, sentences and whole discourses are not interpreted relative to the world, but to the way we know or believe the world to be. It is this mentalist philosophy which also underlies current theories of discourse production and comprehension (van Dijk & Kintsch 1983). In this framework, then, it is assumed that the (relevant) referents of sentences and texts are to be represented as knowledge structures in episodic memory, that is as models (Johnson-Laird 1983; van Dijk & Kintsch 1983; van Dijk 1987b). The structures of these models are not arbitrary. On the contrary, we assume that they have a strategically useful, schematic nature, and consist of hierarchically ordered categories that are always used to 'analyze' real world episodes, a task we accomplish thousands of times each day and which therefore is highly automated.

38

Teun A. van Dijk

There are several reasons to assume that such categories in model-schemata also involve functional categories such as Time, Location, Circumstances, Agent, Patient, Event/Action, etc. These categories not only show up in the semantic role-functions of functional sentence grammars, but also in the categories of storytelling and, indeed, in any verbal 'description' of the world-episode as represented in the model. Hence, the well-known semantic functions that organize the meaning of sentences are derived from, and in fact explained by, the underlying function categories of episodic models. From these cognitive models, which also feature other knowledge (which need not be expressed in the text and therefore can be presupposed), propositions may be selected for communication and expression. In this way a cognitive representation of the meanings of the whole text and its sentences may be strategically built up, and be strategically formulated ('on line') in lexical, syntactic and phonological structures. Obviously, as is well-known in functional and typological grammars, culturally different ways to interpret the (episodes of the) world, presuppose different model-schemata, and hence different ways to organize the semantic structure of sentences, and hence a different syntax (see, e.g., Pawley 1987). 2.2.

Model structures and functional relations in discourse

The next question we may ask is whether also the functional relations between propositions in a text have a cognitive basis. The trivial answer to this question is that this is of course the case for the semantic representation of the text in episodic memory. It is this representation which strategically monitors the actual production of the text, both at the overall macrolevel of text topics, as well as at the microlevel of actually expressed proposition sequences. Less trivial and much more complicated is the question how this representation of the text relates to its 'underlying' models. If a model is a cognitive representation of an episode, it may be assumed to have at least some structures that are analogical to those 'in the world' (Johnson-Laird 1983). This means that for instance spatial, temporal or causal ordering of situations or episodes is mapped into similar, but of course mentally coded, forms of organization in the model. However, discourse is linear and hence requires its own mappings of these possibly 'analogue' model structures. Thus, we speak of a natural (or iconic) order if there is a direct, linear mapping between events represented the model, and those represented and expressed by the text, for instance, first (causing) events precede later (caused) events, as is the case in chronological stories (Chafe 1980, 1987; see also: Dik 1989; Givön 1989).

Issues in Functional Discourse Analysis

39

However, for pragmatic, communicative or interactional reasons, discourse may feature various kinds of transformation of the structures of models. Thus, as we have seen in several examples given above, it may be relevant to present earlier events later in the text, for instance in Explanations. The knowledge that tells us the discrepancies between text and model ordering, as is the case for Explanations, is derived from the model. Besides linear ordering, however, models also feature hierarchical orderings. Information at lower levels of representation, thus, may be subsumed at higher, more abstract levels. For instance, our model of a bank robbery, obtained from actual observation or from a newspaper story, not only may feature such local propositions as 'The robber said: This is a hold-up', but also higher level propositions defining larger action episodes or the event as a whole, e.g. 'The robbers took one million dollars' or, indeed, "There was a hold-up of a bank'. These macropropositions of the model, when relevant for the communicative purposes of the speaker/writer, may also need to be expressed in the text. This means that they must be mapped onto a linear structure of subsequence sentences. Functional relations in discourse precisely allow such mappings. Thus, the Specification relation expresses the hierarchical relation in the model between higher level and lower level information, as in 'John is sick. He has the flu'. Hence, one strategy of model realization is to map top-down relations in the model onto left-right relations in discourse. Apparently, there are general discursive strategies for the linearization of cognitively represented information, for instance in spatial and narrative descriptions (Jarvella & Klein 1982; Levelt 1982). Thus, earlier, larger, closer, more prominent or more important objects or events usually are presented first, as is typically the case in newspaper stories. There are interesting variations of, and exceptions to this strategy, however. Whereas in natural narratives chronological ordering may be dominant, newspaper stories primarily show top-down, relevance ordering: In principle, the most important information comes first (van Dijk 1988). At the level of local text organization, this means that there will primarily be Specification relations. In scholarly discourse, logical relations may be mapped in such a way that inferences tend to come later in the text, as is characteristically the case in the functional semantic Conclusion relation or in the superstructural Conclusion category of an argumentation schema. As soon as these linearization strategies seem to be violated, we assume that the order is 'marked' and requires special interpretation. Thus, sometimes, prominent

40

Teun A. van Dijk

Conclusions may appear first, as also Consequences in linear ordering may be expressed first. We see that the fundamental properties of discourse coherence are textual 'codings' of the ways people organize their mental models about episodes: Temporal, causal, spatial, conceptual or logical relations in the model all have their own mapping between model and semantic representation, and hence between the model and the realization of this representation in the subsequent sentences of the text (see also the contributions in Tomlin 1987a; e.g., Chafe 1987; Dixon 1987; Givön 1987; Tomlin 1987b). The overall linearization strategy in our culture seems to be that old, first, close, recent, left-hand, causing, given, prominent, important, big objects, persons, events or information precede new, later, past, distant, right-hand, unimportant, detailed, small, or following ones. Such relations also underlie the well-known but elusive notion of foreground vs. background information (Givön 1987). However, since there is only one ordering dimension in discourse (in addition to special intonation and graphical expression variation), these mappings may conflict. Earlier events may be less important, and it will depend on the discourse genre whether the realization of the underlying episodic model in that case will be primarily chronological, as in natural stories (although these are usually preceded also by an overall abstract, which at least gives the most important information, and hence the interesting point of the story), or whether the most important information is given first, as is the case in news stories. On the other hand, the important conclusion of an argument may often be mentioned in last position. Similarly, within sentence structure of many languages, 'given', 'old' or other topical information often precedes new, focused and hence more important information. More generally, then, some of the ordering principles of underlying models, e.g., importance, prominence or recency, may be variously mapped as first or as last information in sentence and discourse structures. One of the important factors that also conditions this variation is not the structure of situation models the discourse 'is about', but the structures of the model of the present communicative situation itself, the so-called context model (van Dijk & Kintsch 1983). Thus, the discourse mapping of the information from the situation model is also a function of pragmatic and interactional conditions of the discourse, such as the goals, interests, or the mutual knowledge of the speech participants and the communicative relevance of specific elements of information or discursive (inter)action. For instance, facts or objects may be important, given, close or recent in the situation model, but still occupy a later or less prominent position in discourse for tactical reasons, such as those of face keeping, positive

Issues in Functional Discourse Analysis

41

self-presentation or persuasion. Even esthetic reasons may be involved, such as the convention of revealing the identity of the murderer in traditional crime stories only at the end of the story. Note that context models are also crucial for the description and explanation of meta-linguistic, meta-discursive or meta-communicative information in discourse. Although situation models and communicative context models are the basis of discourse, not all information in discourse derives from models. Indeed, the semantic function of a Generalization may feature information that is not specific to a single episode, but derives from more general knowledge, e.g., as represented in scripts (Schank & Abelson 1977), or other social cognitions, such as general, socially or culturally shared knowledge and beliefs, including attitudes, norms and values, represented in semantic (or social) memory. Stories, thus, may feature generalizations about persons, actions or events. Thus, in racist talk, stereotypes or prejudices may be expressed in order to explain why specific story protagonists, such as ethnic minority group members, act in a specific way (van Dijk 1987a). Similarly, scholarly discourse, and generally argumentation, may feature general norms, rules, or lawlike generalizations, either as conclusions or as part of their Premises (van Eemeren, Grootendorst & Kruiger 1984). Depending on the discourse genre, these may again be mapped linearly by having such general forms of information precede or follow the more specific, instantiated information of the model. From these observations we may generally conclude that global and local discourse ordering, the functional relations between propositions or sentences, as well as the schematic categories of text genres, reflect the underlying cognitive strategies applied in mapping information from episodic models and general social cognitions onto the semantic representation of the text in memory. However, as yet we have little insight into the details of the cognitive representations and the mental strategies involved in these mapping operations. Nor do we know how different mapping principles interact, counter-act or combine with each other in the linearization of cognitive information in semantic representations and 'surface structures' of text and talk.

42 3.

Teun A. van Dijk Social and cultural functions

Having explored some of the cognitive implications, and indeed explanations, of functional categories in sentence and discourse structures, I finally turn to the social and cultural context of both the textual and the cognitive organization and strategies of discourse. At this level of analysis, we first of all deal with discourse as a form of interaction, viz., as a communicative event, which in turn is embedded in more encompassing societal, political or cultural structures. We have encountered some functional aspects of discourse above that are defined in terms of interaction structures. For instance, the moves of a discursive strategy are defined in terms of the goals of the speaker, and hence in terms of properties of interaction. That is, a move is a functional action component which contributes to the realization of the goal of that action (see also Kreckel 1981). Thus, buying a ticket and taking a seat are functional moves in the (scripted) action of going to the movies or taking the train, whereas reading a book on the train is not a functional element of the action of taking the train, simply because without doing so one also realizes the goal of the action. However, it may be a functional move in the strategy of making a pleasant train journey. Similarly, we have seen that Apparent Denials and Apparent Concessions are typical discourse functions in the overall strategy of positive self-presentation, for instance in prejudiced talk. Thus, people do a lot of things *with words', and often they do them at the same time, that is, by doing other things. This parallelism of discourse and interaction also obtains for the functional relations or categories of each level of analysis. This holds for intonation, syntax, semantics, rhetoric, style, superstructures and illocutionary acts, for the (other) interactional and social acts thus accomplished, as well as for their underlying cognitions. Thus, a negation followed by a contrastive proposition, functioning as a Denial, not only strategically contributes to the interactional goal of avoiding a bad impression (Arkin 1981; Brown & Levinson 1987), thereby managing the model the hearer is making of the speaker. Such a denial also functions socio-culturally: By denying that they have something against 'foreigners', speakers also emphasize that they know the official norms of tolerance of their group or culture, while at the same time asserting group membership and confirming ingroup bonds by presenting others in a negative light. Theoretically, however, we need to distinguish in this case between what may be called vertical and horizontal functionality. A proposition functioning as a pragmatic denial, and a pragmatic denial functioning as a socio-cultural act of group cohesiveness or ethnocentrism, exhibit vertical functions between different

Issues in Functional Discourse Analysis

43

levels of analysis. This kind of functionality obtains when we refer to the cognitive, pragmatic, or social 'functions' of language. Such functions may be described as inter-level mapping operations, for instance between model structures and semantic structures of sentences and discourse. Also, they may be characterized in terms of interdependencies and constraints. This is the case when communicative goals limit or prescribe the choices or variation of lexical, syntactic, semantic, stylistic or pragmatic structures of text and talk, for instance when politeness strategies call for the selection of 'polite' language forms, or when formal situations require 'formal' language style. Grammatical, discursive and interactional functions, however, are horizontal or linear, and obtain between units of the same level, e.g., between propositions, between subsequent acts, between propositions and the text as a whole, or between partial acts and the overall action as a whole. Besides the usual ordering relations of precedence and consequence, we find here the functional relations of Generalization/Specification mapping the hierarchical conceptualization of models in linear discourse structures, but also the relations of identity/difference (as in Repetition functions), of whole/part, set/element, old/new, given/derived, cause/consequence, introduction/conclusion, beginning/end, etc. Most of these functional relations hold both for proposition and for action sequences, and therefore also organize the temporal, spatial, informational, conceptual or logical structures of situations or episodes. And in the same way as propositions, at the microlevel of text analysis, may have functions relative to the text as whole at the macrolevel, situations and their structures at the microlevel may have functions at the macrolevel of societal and cultural structure. Thus, the strategic move of a negative presentation of outgroup members in discourse, is a functional element of the more complex act of the reproduction of ethnic prejudice, as well as of the socio-cultural and political strategy of outgroup exclusion and marginalization characterizing white group dominance. It hardly needs to be emphasized that such broader socio-cultural strategies need not be intended, as such. Indeed, as white ingroup speakers emphatically repeat, their negative remarks or even whole biased stories about foreigners are precisely not intended as an element of a pattern or as a step in a strategy, but as a local incident, as a well-motivated exception, lest the interlocutor does make the feared fatal inference: He is a racist. However, whereas for the anti-racist such Apparent Denials may function as a discursive signal which suggests that precisely such an inference may well be correct, for those who share the negative attitude about the outgroup it may function as a social warrant of good citizenship, viz., that one may at the same time respect the

44

Teun A. van Dijk

official norms of tolerance and yet not like 'those foreigners'. In other words, denials and similar discursive and interactional moves both function as the resolution of cognitive inconsistency and as the resolution of a social predicament. We see that semantic, pragmatic or rhetorical discourse functions are related in many ways with interactional and socio-cultural functions. However, such relations are not direct. They always require a socio-cognitive interface, that is, episodic models of specific socio-cultural events and shared general knowledge or attitudes about societal structures, which both can be mapped on, or constrain the cognitive representations that underlie a specific discourse. The functionality of discourse and its elements at various levels of description derives from these cognitive strategies. It shows how the many spatial, temporal, logical or conceptual relations of models are translated into relations between propositions or speech acts linearly expressed by the sentences of a text or dialogue. Functionality principles also show how both discourse and such cognitions are in turn embedded in strategic interaction and within a broader structure of socio-cultural reproduction, an embeddedness which again is cognitively represented, viz., in terms of the relations between particular models and the general knowledge, attitudes and ideologies of shared social cognitions of the group. Indeed, selectively hearing, remembering and retelling a story about 'those Turks down the street' strategically, and hence functionally, instantiates in a narratively expressed model a more general attitude and ideology about 'foreigners', a social cognition that happens to function itself in the reproduction of white group power. Hence functionality in language use and discourse not only shows how sentences or texts are organized, and not only how they are related with cognition, interaction and social structure. It also shows that discourse is never context-free, and hence never innocent.

Issues in Functional Discourse Analysis

45

REFERENCES Arkin, R. M. 1981. Self-presentation styles. In: J. T. Tedeschi (ed.) Impression management. Theory and social psychological research. 311-333. New York: Academic Press. Bolkestein, A. M., de Groot, C., & Mackenzie, J. L. (eds.) 1985. Syntax and pragmatics in functional grammar. Dordrecht: Foris. Brown, P., & Levinson, S. C. 1987. Politeness: Some universal in language usage. Cambridge: Cambridge University Press. Chafe, W. L. (ed.) 1980. The pear stories. Hillsdale, NJ: Erlbaum. Chafe, W. L. 1987. Cognitive constraints on information flow. In: R. Tomlin (ed.) Coherence and grounding in discourse 21-52. Amsterdam: Benjamins. DaneS, F. (ed.) 1974. Papers on Functional Sentence Perspective. The Hague: Mouton. Dik, S. C. 1978. Functional grammar. Amsterdam: North Holland. Dik, S. C. 1980. Studies in functional grammar. London: Academic Press. Dik, S. C. (ed.) 1983. Advances in functional grammar. Dordrecht: Foris. Dik, S. C. 1989. The theory of functional grammar. Part I: The structure of the clause. Dordrecht: Foris. Dixon, P. 1987. Actions and procedural directions. In: R. S. Tomlin (ed.) Coherence and grounding in discourse, 69-89. Amsterdam: Benjamins. Fillmore, C. J. 1968. The case for case. In: E. Bach, & R. T. Harms (eds.) Universals in linguistic theory, 1-88. New York: Holt, Rinehart & Winston. Givön, Τ. 1979. On understanding grammar. New York: Academic Press. Givön, Τ. (ed.) 1983. Topic continuity in discourse: Quantitative cross-language studies. Amsterdam: Benjamins. Giv6n, T. 1987. Beyond foreground and background. In: R. S. Tomlin (ed.) Coherence and pounding in discourse, 175-188. Amsterdam: Benjamins. Givön, Τ. 1989a. Mind, code and context: Essays in pragmatics. Hillsdale, NJ: Erlbaum. Givön, Τ. 1989b. The grammar of referential coherence as mental processing instructures. Eugene, OR: University of Oregon, Linguistics Department. Unpublished paper. Grimes, J. E. 1975. The thread of discourse. The Hague: Mouton. Haiman, J., & Thompson, S. A. (eds.) 1988. Clause combining in grammar and discourse. Amsterdam: Benjamins. Halliday, Μ. A. K. 1985. An introduction to functional grammar. London: Edward Arnold. Jarvella, R. J. & Klein, W. (eds.) 1982. Speech, place and action. Chichester: Wiley. Johnson-Laird, P. N. 1983. Mental models. Cambridge: Cambridge University Press.

46

Teun A. van Dijk

Kreckel, M. 1981. Communicative acts and shared knowledge in natural discourse. London: Academic Press. Levelt, W. J. M. 1982. Linearization in describing spatial networks. In: S. Peters, & E. Saarinen (eds.) Processes, beliefs and questions. Dordrecht: Reidel. Longacre, R. (ed.) 1977. Discourse grammar. 3 vols. Dallas, TX: Summer Institute of Linguistics. Mann, W. C. & Thompson, S. A. 1988. Rhetorical Structure Theory: Towards a functional theory of text organization. Text 8, 243-281. Mann, W. C. Matthiessen, C. Μ. I. M., Thompson, S. A. 1989. Rhetorical Structure Theory and text analysis. Los Angeles: University of Southern California, Information Sciences Institute. Research Report. Meyer, B. J. F. 1975. The organization of prose and its effects on memory. Amsterdam: North Holland. Pawley, A. 1987. Encoding events in Kalam and English: Different logics for reporting experience. In R. Tomlin (ed.) Coherence and grounding in discourse 329-360). Amsterdam: Benjamins. Pike, K. L. 1967. Language in relation to a unified theory of human behavior. The Hague: Mouton. Schänk, R. C., & Abelson, R. P. 1977. Scripts, plans, goals and understanding. Hillsdale, NJ: Erlbaum. Steiner, Ε. Η., & Veltman, R. (eds.) 1988. Pragmatics, discourse and text. Some systemically inspired approaches. London: Pinter. Tomlin, R. S. (ed.) 1987a. Coherence and grounding in discourse. Amsterdam: Benjamins. Tomlin, R. S. 1987b. Linguistic relections of cognitive events. In: R.S. Tomlin, (ed.) Coherence and grounding in discourse. 455-480). Amsterdam: Benjamins. van Dijk, T. A. 1977. Text and context. London: Longman. van Dijk, T. A. 1980. Macrostmctures. Hillsdale, NJ: Erlbaum. van Dijk, T. A. 1981. Studies in the pragmatics of discourse. The Hague/Berlin: Mouton. van Dijk, T. A. 1984. Prejudice in discourse. Amsterdam: Benjamins. van Dijk, T. A. 1987a. Communicating Racism. Ethnic Prejudice in Thought and Talk. Newbury Park, CA: Sage. van Dijk, T. A. 1987b. Episodic models in discourse processing. 1983. In: R. Horowitz, & S. J. Samuels (eds.) Comprehending oral and written language, pp. 161-196. New York: Academic Press. van Dijk, T. A. 1988. News as Discourse. Hillsdale, NJ: Erlbaum. van Dijk, T. A. & Kintsch, W. 1983. Strategies of discourse comprehension. New York: Academic Press.

Intransitivity, transitivity and control in Lillooet Salish

Jan P. van Eijk Saskatchewan Indian Federated College Canada

0.

Introduction

This article discusses the intersecting categories Intransitive, Transitive and Control in Lillooet Salish. The article is based on research that was first presented in the author's doctoral dissertation (Van Eijk 1985), but it also incorporates corrections and new insights that have resulted from subsequent research. The final section of the article discusses intransitivity, transitivity and control in the context of the larger Salish picture, and within the theoretical framework of Functional Grammar as developed by Simon Dik.

1.

Lillooet: general overview

Lillooet is an Interior Salish language spoken in British Columbia, Canada.1 Lillooet phonemes are the following: ρ t c ς k kw q qw p' c' k' k' w q' q' w 4 s s w χ x χ m m' η η' 1 1' } }' y y* ζ ζ' γ γ ' S ϊ ' i , w w w' h ? a q i j u i j a ? . (The superscript comma indicates glottalization, while the subscript dot indicates retraction, i.e., velarization and simultaneous tensing. For further phonetic and phonological details I refer to Van Eijk 1985,1988.) Morphologically, Lillooet words fall into two basic classes: variable and nonvariable words. A variable word consists of a stem which in itself consists of a root

48

Jan P. van Eijk

to which one or more reduplicative or non-reduplicative affixes may be added.2 To such a stem we can add an outer layer of so-called personal affixes, which express possession ('my', 'his', etc.), subject (T, 'he', etc.), or object ('me', 'him', etc.).3 Personal affixes distinguish three persons in singular and plural. Stem-forming (i.e., non-personal) affixes express a variety of notions (lexical, aspectual, nominalizing, verbalizing, abstract/aspectoid, reflexive, reciprocal, 4 intransitivizing and transitivizing). Of immediate interest to our study are the intransitivizing and transitivizing affixes ('intransitivizers' and 'transitivizers' for short), all of which are non-reduplicative suffixes.

2.

Intransitive and transitive stems

Lillooet stems can be exhaustively divided into two classes: intransitive and transitive. Transitive stems (all verbs) are all marked with an overt transitivizer, and these stems are the only ones which can combine with object suffixes and with reflexive and reciprocal suffixes. Intransitive stems (which fall into verbs and nouns) 5 do not combine with object suffixes or with reflexive or reciprocal suffixes. To give an example, from transitive nuk'^-an 'to help s.o.' (-an transitivizer) we form nuk' w ?-an-ci4kan Ί (4kan) helped you (-ci object suffix 2S)', but from intransitive nuk'w2-am 'to be of help, to provide help' (-am intransitivizer) we cannot form *nuk""?-am-ci-lkan. Intransitive verbs include action terms (such as ntfk,w?-am above), numerals (e.g., päla? '(to be) one') and adjectives (e.g., TdlSal '(to be) strong'). 2.1.

Intransitivizers

Lillooet intransitivizers (i.e., the markers of intransitive verbs) fall into three types: zero, M, -xal. Zero intransitivizers comprise verbs without an overt intransitive marker, e.g., li'iq 'to arrive (here)', qam't 'to get hit', qäm't-ac 'to get hit on the mouth (-ac)', and päla? and SälSal (see section 2 above). Μ intransitivizers fall into various formal subtypes, viz., -am, -am, -im, -um (in some cases also with m glottalized). The intransitivizer -xal is not divided into subtypes. Virtually all cases with -xal, and a number of those with Μ or zero refer to an implied object, e.g., püi-xal 'to boil things' (French faire bouillir), k'äx-xal 'to dry things' (French faire sicher ), nik'-xal 'to cut things', q' w al-xäl 'to roast, cook things', q'w31-am 'to roast, cook things' (id. as q' w al-xäl), ?äc'i-am 'to see (things)', ndk'w?-am 'to be of help, to provide help', päi-alq w -am 'to scrape (ρά£-) logs (-alq w )', ?dqwa? 'to drink', c'aq' 'to throw'. These 'object-implying' intransitives usually also combine with adjuncts which translate as object complements, e.g., wa? k'äx-xal ki-maxäz'-a 'she is drying huckleberries (maxäz')'. 6 However, since these verbs may not combine with object suffixes (or reflexive or reciprocal suffixes) they are not transitive from a Lillooet point of view. Intransitive verbs that do not refer to an implied object are passive or medialreflexive in character. Here belong the remaining cases with M, and almost all unmarked stems, e.g., li'iq, qam't, qäm't-ac, päla?, iälSal (see above), k' w zus-am

49

Intransitivity, Transitivity and Control in Lillooet Salish

'to work', ptinus-am 'to think'. (In particular, Μ intransitivizers refer to 'one's own body part' when occurring after somatic suffixes - a subset of the lexical suffixes - as in sup-akä?-am 'to scratch (sup-) one's hand (-akä?)'.) The relationships between the various types of intransitive and transitive stems can be read from the following chart:7 Stems

Transitives (Verbs)

Intransitives

Verbs

Marked (-M, -xal)

Nouns

Unmarked zero

The intransitivizer -am combines with nouns into forms which express 'to hunt, pick, obtain the referent of the noun'. (The nouns are then verbalized by -am.) Examples are: s-iiq 'duck' (s- nominalizing prefix) —> xiq-am 'to hunt duck', s-p'ams 'firewood' —> p'äms-am 'to collect firewood'.8 2.2.

Transitivizers

The transitivizers fall into four functional types which are tabulated below. Comments follow. direct

indirect

plain

-s, -an-s -nun/-nun\ Ν

-xit

relational

-min/-min'

-min-xit/-min'-xit

The transitivizers -s and -an-s require formally different object and subject suffixes for certain persons than the other transitivizers. For example, besides ^'iq-s-tum'x-as 'he (-as) brought Gi'iq-s) me (-tum'x)' we have ?äc'x-3n-c-as 'he saw (?äc'x-an) me (-c)' 'to arrive (here)', ?ac'x 'to be seen'; for -an (an allomorph of the Ν transitivizer) see below). The transitivizer Ν falls into a number of formal subtypes (-an, -an', etc.), the distribution between which is partially predicable on the basis of the formal or semantic structure of the preceding stem. For example, many roots CAC (where A is a a, i i. or u y) take -An', where A duplicates the vowel of the root, e.g., p&-un'

50

Jan P. van Eijk

'to boil s.t.', k'äx-an' 'to dry s.t.', nik'-in' 'to cut s.t.' (Compare these cases with pui-xal, k'äx-xal and nik'-xal in 2.1.) The transitivizer -an-s consists of the transitivizer -s (see *i'iq-s above) plus an element -an which may be the transitivizer -an (see ?äc'x-an above) or a different suffix with an unclear semantic function. The transitivizer -min-xit (-min'-xit) is a combination of -min (-min')with the transitivizer -xit. The distribution between -nun and -nun', and between -min and -min' (and -minxit and min'-xit) is largely predictable in terms of the formal structure of the preceding stem. However, a detailed discussion of this matter falls outside the scope of this article. The functions of the various transitivizers are discussed in the next two sections. We discuss plain-direct intransitivizers in 3, and the remaining transitivizers (plainindirect, relational-direct and relational-indirect) in 4.

3.

Transitivizers: the plain-direct class

The transitivizer -s serves three functions: (1) expressing causativity, as in^i'iq 'to arrive (here)' --> iCiq-s ' to bring s.o. or s.t. (= to cause to arrive here)', qam't ' to get hit' — > qam't-s 'to hit s.o. or s.t.', n'as 'to go' —> n'as-c 'to take s.o. or s.t. along', xä 'to work out, to get done' —> xä -c 'to do s.t.' (as the last two examples show, -s has the shape -c after stems ending in s or4); (2) forming transitive verba declarandi, which imply an adressee, as in qwal'Ut 'to speak, hold a speech' --> qwaltft-s 'to speak to s.o., admonish or rebuke s.o.', x'itan 'to whistle' --> xwitan-s 'to whistle at s.o.'; (3) forming transitive verba sentiendi which imply that the subject nourishes a certain thought on the object, as in xzum 'big, great' —> xzum-s 'to respect s.o.', «i'ax 'sweet, tasty' —> Ä'sx-s 'to like s.t. (i.e., food).' The functions (1) and (2) are also performed by Ν transitivizers: (1) expression of causativity: k'ax '(to be) dry' --> k'ax-an' 'to dry s.t.', ?ac'x 'to be seen' --> ?äc'xan 'to see s.o. or s.t.', xzum 'big, great' —> izüm-un 'to make s.t. bigger' (cf. this last example with xzum-s above);9 (2) forming transitive verba declarandi: sqwal' 'to report, bring news' —> sqw£l'-an 'to report to s.o.', n-q"al'ut-äna? 'to hint, drop hints' (q w al'ut 'to speak', n-...-ana? 'ear') --> n-qwal'ut-£n?-an 'to hint to s.o.' (with regular shortening of -ana? to -an?), p t a k l 'to tell a legend' --> ptäkl-an 'to tell a legend to s.o.', wa?äw 'to shout, holler', --> wa?£w-an 'to shout at s.o.'. Function (3), forming transitive verba sentiendi, is also performed by -nun/-nun' (but not by N), as in wanäx" 'true' --> wanäxw-nun' 'to believe s.o.', käkza? 'to lie, tell a lie' --> kakza?-ntfn 'to doubt s.o.', x™?uxw 'to smell' --> χ*?ύχ*-ηιιη' 'to smell s.t.'.10 There is also a fourth function, viz., direct (non-causative) transitivizations of verbs that are not verba sentiendi or declarandi. This function is served by Ν transitivizers in combination with certain stems, and by the transitivizer -an-s. With Ν transitivizers we have first of all the cases c'aq' 'to throw' - > c'aq'-an 'to throw at s.o., to hit s.o. by throwing' (literally 'to bethrow s.o.'), ptak 'to pass by' ptak-an 'to pass by s.t. or s.o.'. In the second place, combinations of nouns and the

Intransitivity, Transitivity and Control inLillooet Salish

51

transitivizer -an express 'to infest with the referent of the noun, to apply the referent of the noun to s.t.'. These nouns refer to animals, and the transitive forms based on these nouns are usually in the passive, as in pu?y'axw-än-am 'it is infested with mice (puTy'ax*)' (-am passive marker, not to be confused with the intransitivizer -am), makn'-än-am 'it is infested with lice (m^kan')', pul'h-in-am 'it is infested with piil'a (maggots when still in the egg)', k'ak^?-än-am 'it is infested with k'äk'-4a? (bugs in dried salmon)'. (The morphophonemic changes in the last three forms are regular.) The transitivizer -an-s is mostly used on unmarked intransitive stems that refer to an object, as in naq' w 'to steal' —>naq'w-an-s 'to steal s.t.', qan'im 'to hear' —> qan'im-an-s 'to hear s.t.', ? "Mq^-an-s 'to drink s.t.' (with regular morphophonemic adjustments). The criteria which determine when a certain function is served now by one transitivizer, then by another, are discussed in 3.1. and 3.2. 3.1.

Transitivity and control

The difference between -s and Ν with regard to function (1), expression of causativity, is that -s does not express control of the subject, while Ν expresses that the subject is in control. For example, in *i'iq-s 'to bring (here)' (= 'to cause to arrive here'), the subject causes the object to carry out an action over which neither is in control (i.e., neither the entity brought, nor the agent causing the arriving is in complete control over whether the bringing will be accomplished). Similarly, in qam't-s 'to (cause to get) hit', the object is not in control over its being hit, while the subject is not in control over the hitting (one does not know in advance whether or not one will score a hit): hence here we have a total lack of control. However, in k'äx-an' 'to dry s.t.' (= 'to cause it to get dry'), the subject is in total control over the object. (For example, one can put berries in the sun and keep them there until they are dried.) Illustrative are the rare cases where we have both -s and Ν after the same stem, e.g. cukw-s 'to finish s.t.' vs. cukw-un' 'to put the finishing touches on s.t.' (cuk® 'to be finished, to quit'), cam'p-s 'to get to the end of s.t.' vs. cam'p-än 'to work on s.t. until it is finished' (camp 'finished, having gotten to the end'): on may not know in advance when one will finish s.t. (cukw-s) or when one will come to the end of s.t. (cam'p-s), but one is more in control when putting on the finishing touches (cukw-un' ) or when working on s.t. until it is finished (cam'p-än). Note also cukw-un'4kän-kl Ί (4kän) will (-fei) put the finishing touches on it later', besides which *cdkw-s-kan-fel was rejected: although -kl 'remote future' may be combined with verbs with -s, it apparently combines more naturally with Ν where there is a choice between -s and N, since Ν expresses the total control that is neccesary to make statements about future actions the outcome of which is certain. The transitivizer -s may also follow intransitivizers in order to indicate causativity, as in xäii'-xal-s 'to let s.o. take things out' (χ&ί'-χal 'to take s.t. out of a container'), ptinus-am-s 'to make s.o. think' (ptinus-am 'to think'), c'aw'-äk?-am-s 'to let s.o. wash his hands (by getting soap, water and towel ready)' (c'aw'- 'to wash', aka? 'hands', here regularly reduced to -ak?): these cases show shared

52

Jan P. van Eijk

control between subject and object, i.e., the subject is in control over the action, but so is the object. Some affixes mark a stem as being 'non-control'. In other words, these affixes indicate that the protagonist of the action denoted by the stem is not enirely in control of that action. For example, Lillooet employs so-called 'final reduplication', which repeats the consonant after the stressed vowel in a word (and in general separates this consonant from its copy through a). This operation has an aspectual function which can be broadly labeled 'continuative' or 'recursive', but it also indicates non-control (the protagonist is not in control over the final completion or outcome of the process). When stems with final reduplication are causativized, the chosen transitivizer is invariably -s, even if the stem without reduplication has another transitivizer. Examples are: zawit-at-s 'to know s.t., to have learned s.t.' (cf. zawät-an 'to know s.t.;' zawät-at-s indicates that the knowledge was acquired after some trying or a conscious learning effort, in other words, that the protagonist did not have a preestablished control over the knowledge), ptak-ak-M-s 'to send s.o. right by s.t.' (-"Ml 'real, very much, too much'; cf. ptäk-an 'to pass by s.t.'),4ilal-s 'to sprinkle s.t. too much, to overdo it when sprinkling s.t.' (cf. 4il-in 'to sprinkle s.t. (e.g. flowers)'). The combination of the prefix ka- and the enclitic -a (ka-...-a) indicates the aspectual notion 'resultative'. It also indicates that the action referred to by the stem enclasped by ka-...-a is achieved either suddenly or at the cost of some effort, and that the protagonist of the action has a lessened control over it. When stems with an Ν transitivizer are combined with the ka-...-a combination, the Ν transitivizer is replaced with -s, as in ?äc'x-an 'to see s.o. or s.t.' —> ka-?äc'x-s-a 'to catch sight o f , t3q-an 'to touch s.o. or s.t.' --> ka-täq-s-a 'to catch s.t. (e.g., a ball)', ?w31-an 'to light s.t., set fire to it' --> ka-iw31-s-a- 'to manage to light it'. In the above cases, -s is used only where we do not have full control of the subject over the action. However, -s is not a 'non-control' transitivizer, but rather it is indifferent (or neutral) with regard to control. On the other hand, Ν is definitely used to mark full control of the subject over the action. The neutral function of -s with regard to control is demonstrated by the fact that when stems with an Ν transitivizer are combined with the stative prefix s- (not to be confused with the nominalizer sdiscussed in 2.1), the Ν transitivizer is replaced with -s. (The stative prefix s- marks a state resulting from an action. It also indicates that the performer of that action is in control over the action and the object.) Examples are: s-?adi-s 'to watch over, look after', s-teq-s 'to hold on to', s-iwal-s 'to keep lit' (see the preceding paragraph for the forms with N). Note that stative s- implies full control, while ka-...-a implies lack of control. Hence when a stem is marked for control (either positively or negatively) we have -s as transitivizer (i.e., the 'neutral control' transitivizer). Summing up the above facts with regard to control, we arrive at the following schema (X refers to 'stem'):

53

Intransitivity, Transitivity and Control in Lillooet Salish

control stem unmarked for control

X-N

stem marked for control

s-X-s

no control X-s ka-X-s-a

Stems with the transitivizer -s change -s to -an before the reflexive suffix -ciSt, as in i{'ak-s 'to take along1 (li'ak 'to go') —> ii'ak-an-cut 'to take oneself along, to take care of one's own transportation', n'as-c 'to bring along' (n'as 'to go') —> n'as-anctft 'to take care of one's own transportation', qam't-s 'to hit (qam't 'to get hit') —> qam't-an-ctft 'to hit oneself, zuqw-s 'to kill (an animal), to beat s.o. up', (zuqw 'to die') —> zuq w -an-cut 'to kill oneself. The suffix -cdt 'de-marks' the stem with regard to control, and -an is chosen before -cut to indicate that the subject is in control of the action. Stems with -an-cut may be secondarily marked for control by ka-...-a or s-, as in s-?ac'x-an-cut 'to look after oneself, ka-ii'ak-an-ctft-a 'to manage to provide one's own transportation'. Note that s- and ka-...-a only operate on transitivizers that conclude a stem and that they do not influence the choice of Ν before -cut. 3.2.

Transitivity and control: non-causative cases

Where we have non-causative transitivizations, i.e. in functions (2)-(4), it is not clear on what basis one or the other transitivizer is selected in order to serve a particular function. For example, in function (2), expression of transitive verba declarandi, the selection of -s vs. Ν seems unpredictable on the basis of the semantics of the stem. (Remarkably, qwal'V)t-s 'to speak to' and sq w äl'-3n 'to report to' (section 3) both contain the root q w al'- 'to speak'. Compare also q w al'iJt-s with n-q w al'tft-4n?-an 'to hint to', section 3.) It is possible that again control is a factor in the choice between s and N. (It might be that from a Lillooet point of view cases with Ν suggest that the message does get across to the addressee, while -s might suggest that the message may or may not get across.) It is equally unclear why function (3), the formation of transitive verba sentiendi, is in some cases expressed by -nun/-nun\ in others by -s (although the cases with -s seem to suggest a greater degree of subjectivity: it is a moot point whether someone is worthy of respect (Xzum-s 'to respect') but one can test objectively whether someone is to be believed (wanäx w -nun' 'to believe')). In the same way, there does not seem to be an objective criterion for predicting the distribution between Ν and -an-s in function (4). The affixes that mark or de-mark a stem for control (ka-...-a, s-, -ctft) are rarely found in combination with stems that serve functions (2) through (4). This lacuna probably results from an accidental gap in the corpus, rather than form a bar on such combinations. However, we do have a number of cases with -s in function (3) combined with -cut, viz., sama?-s-cut 'to pretend to be a white person (sima?)', k w uk w pey'-s-cut 'to pretend to be a chief (k w uk w pey'), to be bossy', laqsay^-ctft 'to

54

Jan P. van Eijk

want to be petted, to think one is the pet (lsqsäy'4) more than others' (the dropping of -s between 4 and -cdt is regular), k' w uk' w m'it-s-cdt 'to act like a child (sk' w uk' w m'it)', xa?-s-cdt 'to brag' (xa? 'high'), ialial-s-cut 'to want to be tough' (Ϊ3ΐ?3ΐ 'stong'). Note that here, in contrast to cases where -cut combines with a stem in function (1), we do not have a switch from -s to another transitivizer (which in function (3) would be -nun/-nun')· We also have a few cases of Ν in function (4) before - c d t , viz. qayx w -än-cut 'to act like a man, to do something in a man's way' (referring to a man doing woman's work in a clumsy manner; s-qayx w 'man'), yaqc?-än'cut 'to do a man's job, but without being good at it' (said about a woman; s-yäqca?- 'woman'), k' w uk' w m'it-än-cut 'to act like a child' (according to my consultants, this word is practically identical in meaning to k' w uk' w m'it-s-cdt), TalSal-än-cut 'to act tough, to want to be brave (when one is not), to force oneself to do s.t.' (cf. Talial-s-cdt above), qglqaj-xn-^n-cut 'to suffer from lack of help; qaj 'bad', -xn- 'foot, leg' (also used in a number of words referring to having trouble), S w 3st-än-cut 'to make a big effort' (cf. x w ux"st 'to need s.t.').11 However, the cases with -än-cut are not indicative for the control status of Ν in function (4), since Ν is the expected transitivizer before -cdt (based on the situation in function (1), see ji'akan-cdt above) and since we do not have examples of these stems with a transitivizer but without -cdt. The plain-direct transitivizers and their functions are represented in the following chart. The dotted line indicates that the criterion 'control' - 'neutral control' is possibly invalid for functions (2) through (4).

control

neutral control

(1) causativizing

Ν

-s

(2) referring to addressee 0verba declarandi) (3) expressing a thought (verba sentiendi) (4) non-causative transit!vization

Ν

-s

-nun/-nun'

-s

Ν

-sn-s

4.

Relational and indirect transitivizers

The difference between direct (plain and relational) and the indirect (plain and relational) transitivizers is that direct transitivizers form two-place verbs, while indirect transitivizers form three-place verbs. In other words, verbs with direct transitivizers imply an object and a subject, while verbs with an indirect transitivizer imply an indirect object in addition to a (direct) object and a subject. The difference between plain and relational transitivizers is that plain transitivizers refer to an object that is affected itself by the activity implied by the stem, while

55

Intransitivity, Transitivity and Control in Lillooet Salish

relational transitivizers refer to an object that is not affected itself, but only involved in the action in an oblique manner. Typically, relational transitivizers refer to an object with or about which on does something, rather than to an object to which one does something. The relational-indirect transitivizer (-min-xit/-min'-xit) combines both the relational and indirect notions in order to express 'to do s.t. for s.o. with regard to that person's possessions'. Below, we list a number of examples with relational and indirect transitivizers, contrasted to cases with plain-direct transitivizers. (A)

Relational-direct transitivizers.

?i?wa?-min p'än't-min' kwtflan-min päqwu?-min ntfk'wa?-min Tul'l'us-min (Wl'l'us ptäkl-min sq w äl'-min' ?üqwa?-min c'äq'-min' täw-min'

'to go along (2i?wa?) with' 'to return (p'an't) for s.t. to return to get s.t.' 'to borrow (kwiJian) s.t. from s.o.' 'to be afraid (päqwu?) of 'to accompany, go with' 'to join, go with s.o.' 'to get together) 'to tell a legend (peak1*!) about s.o.' 'to report on s.o.' 'to drink (Wqwa?) s.t. away, to lose by drinking' 'to throw (c'äq') s.t.' 'to sell (taw-) s.t.'

2?wa?-s p'an't-s

'to take s.o. along' 'to return s.t.'

kw&3n-s

'to lend s.t. to s.o.'

päq^-an nük'^-an TulVus-an pt£k*l-an

'to scare s.o.' 'to help s.o.' 'to gather, bring together' 'to tell a legend to s.o.

sq w äl'-an Wqw?-an-s

'to report to s.o.' 'to drink s.t. up'

c'äq'-an täw-an

'to throw s.t. at s.o.' 'to sell s.t. to s.o.'

A case like päqwu?-min/päq"?-an is typical: a person whom one fears (päqwu?-min) is not affected by this fear, but a person whom one scares (päqw?-an) is affected in that he has fear put into him. In the same way, a person or animal about whom one tells a legend (ptäk^-min) is not affected by the telling of the legend, but a person to whom one tells a legend (ptäM-an) is affected in that he receives information contained in the legend. When combined with -min, nouns express various notions. (Combinations of -min' with nouns were not recorded.) Examples are: stqin-min 'to use s.t. for a pillow (stqin)', c'äz'q-min 'to use s.t. for a cushion (c'az'q)', szäytan-min 'to carry out, to execute s.t.' (szäytan 'business, what one does'), saw't-min-am 'he was taken slave' (saw't 'slave'). When combined with names for animals, -min expresses 'animal destroys, eats, meets s.t.' Such forms are usually in the passive. Examples are: mixa4-min-sm 'it was eaten by a bear (mixal), he was met by a bear, ran into a bear', wuw'a-min-am 'it was eaten by a cougar (s-wUw'a)', pu?y'axw-min-3m 'it was eaten by mice (pu7y'axw).

56

(Β)

Jan P. van Eijk

Plain-indirect transitivizers:

ißq-xit cul-xit q w al-xit niq' w -xit

'to 'to 'to 'to

bring s.t. to s.o.' point s.t. out to s.o.' roast, cook s.t. for s.o.' steal s.t. from s.o.'

iPiq-s cul-un' q w 3l-an näq' w -an-s

'to 'to 'to 'to

bring s.t. or s.o.' point at s.t.' roast, cook s.t.' steal s.t.'

As we have said above and as the examples illustrate, -xit refers to an indirect object, i.e., a benificiaiy or recipient, or the one from whom something is taken. The object suffix that follows -xit refers to the indirect object. Predications with -xit may also refer to a direct object, which is then expressed by an object complement, as in ciSxi-c-as ti-citx w -s-a 'he (-as) pointed out his (-s) house (citxw) to me (-c)' (with regular merger of the final t of -xit with -c). 1 2 (C)

Relational-indirect transitivizers:

nPiq-min'-xit c?äs-min'-xit tx w us-min-xit

'to arrive here for s.o.'s possessions' 'to come here for s.o.'s possessions' 'to look out on s.o.'s behalf for s.t. belonging to that person'

The object suffix following -min-xit/-min'-xit refers to the indirect object, while the direct object is expressed by an object complement, as in tx w us-min'-xi-c-kax w ni-nc'qäx?-a 'you (-kax w ) look out for my (n-) horse (c'qäxä?) for me (-c) (so you can tell me where it is or you can bring it to me)', c?as-min'-xit-(7-kan kw-s-Kika? ?ix"ik'äy'-s-a Ί am coming to get the prepared salmon that belongs to Kika? (so I can bring it to Kika?)' (the zero object suffix 0 refers to Kika?; proper nouns are regularly extended with the nominalizer s- and take the article kw-; x w ik'4z' 'prepared salmon', here regularly adjusted to x w ik'iy' before -s 'his, her'). The indirect and relational transitivizers do not distinguish between control and neutral control forms. As a result, they are unaffected by those affixes which mark a stem for control (ka-...-a, S-, final reduplication), e.g., ka4ax w -min-0 4kän-a Ί (4kän) sneaked up (4axw-) to it (β).' Stems with -min/-min' are extended with -an when combined with -cut (which, as we have seen in 3.1., de-marks a stem for control): päqwu?-min 'to be afraid of s.t.' --> päq w u?-min-an-cdt 'to be afraid of oneself, nk' w z-änwas-min 'to worry about' ~> nk' w z-änwas-min-an-ctit 'to worry about oneself, k'al'än'-min' 'to listen to' —> k'al'än'-min'-an-cut 'to listen to one's own advise', k' w zus-min' 'to work for, to look after' --> k'wzus-min'-an-ci3t 'to look after oneself. The transitivizer -xit does not take an extra transitivizer when combined with -cut: mays-xi-cut 'to fix (mays-) s.t. for oneself', q' w sl-xi-c\it 'to cook (q' w al-) s.t. for oneself, k' w ul'-xi-c\it 'to make (k' w ul'-) s.t. for oneself (in all cases with regular merger of t of -xit with c of -cut).

Intransitivity, Transitivity and Control in Lillooet Salish 5.

57

Intransitivity, transitivity and control: general issues

So far, we have discussed the issue of control in Lillooet with regard to transitive stems. In this section we discuss control with regard to intransitive stems in the context of the larger Salish picture (5.2.), and we relate the issue of control in Salish to recent statements on control within the framework of Functional Grammar (5.3.). In section 5.1. we briefly recapitulate the concept of control. 5.1.

Control: definition and analysis

Having discussed the concept of control in terms of the Lillooet verbal system, it is now propitious to give a concise definition of 'control' which covers not only the Lillooet (and general Salish) facts, but can also be used profitably for non-Salish languages. The best defintion seems to be the one provided by Dik in a recent monograph (Dik 1989). In chapter 5 of his book, Dik discusses 'nuclear predications', which "consist of terms which designate entities in some world, and of predicates, which designate properties of, or relations between such entities. The nuclear predication as a whole designates a set of states of affairs (SoA's), each member of the set being defined by the particular property or relation designated by the predicate' (p. 89). On p. 96, Dik stipulates: "An SoA is [+control] if its first argument has the power to determine whether or not the SoA will obtain". This definition neatly captures the Lillooet facts, and also solves one vexing problem: the Lillooet form for 'to miss (a target), transitive' is xik'-an, with the [+control] transitivizer -an. Similarly, Thompson has xik'-e-t-jO-p 'you people (-p) shot and missed the target', with the control marker -e, the transitivizer -t and the object suffix 3S -0 (Thompson 1985: 414); and Shuswap has xik'-n-s 'he (-s) misses it', also with the control transitivizer -n (Kuijpers 1974: 227). 13 However, as Thompson (op. cit.) observes, the action of 'missing' expresses a lack of control, rather than the presence of control. Dik's definition, with its clear demarcation of the role of the first argument and the outcome of the action, resolves this dilemma: missing a target is entirely within the control of a first argument (in this case, the agent). Hitting a target, however, is outside the agent's control and this is why Lillooet selects -s here (see qam't-s, 3.1.). Thus, 'control' does not express the idea of desiring a goal and then achieving it, but rather the mere ability of a protagonist to achieve a goal, regardless of whether or not that goal is the desired one.14 5.2.

Control in Salish grammar

The category 'control' in Salish grammar has been studied extensively in recent literature. Major contributions on this topic include Thompson 1979b, Thompson 1985 (with a historical survey of the study of 'control' in fn. 3), Carlson and Thompson 1982, and Kinkade 1982. Recent contributions include Kroeber 1988 and Van Eijk (in press). Of these articles, Thompson's beautifully written and closely argued 1985 contribution is the most important, and an indispensible tool for any scholar interested in this concept. Among other things, Thompson pays attention to

58

Jan P. van Eijk

the control status of intransitives. As we have seen in 2.1., we can divide Lillooet intransitives (and Salish intransitives in general) into those stem which imply an object (goal, patient), and those which are passive or medial-reflexive in character. Thompson (pp. 396-8, 401-10) argues persuasively for a different dichotomy, viz., that between non-control intransitives (basically the passive intransitives) and control intransitives (encompassing the medial-reflexive and 'with object' intransitives). In Lillooet, when the non-control intransitives are causativized we have a choice between those where the protagonist has the power to determine whether the goal will be achieved, and which take Ν (k'rfx-an', q'w31-an), and those where the protagonist has no such power, and where -s is selected (qam't-s, ^i'iq-s).15 Where a control intransitive is causativized, we have only the situation where the protagonist does not have complete control, but shares it with the causee, and here the only choice is -s (x&i'-xal-s, ptinus-am-s).16 Since both [+control] intransitive stems and [-control] intransitive stems allow causativizations with -s, a definite criterion for what constitutes a [+control] or a [-control] intransitive stem is not always easy to establish. English translations are, of course, no help. One can translate Lillooet the

Auxiliary Operator Inf and constructions with an Infinitive in Hungarian

81

In this example a trigger Τ introduces a formal element F. However, languages can also have formal elements which cannot be considered the immediate expression of a trigger. In some cases a trigger introduces not a form, but some other trigger. The distinction between the initial trigger and the second trigger may be indicated by the terms 'primary versus secondary triggers' (De Groot 1990) or 'primary versus auxiliary operators' (Dik 1989). We will adopt the terminology used by Dik (1989). It is important to distinguish between primary and auxiliary operators because the occurrence of certain formal elements depends on other grammatical devices. In our opinion, this is illucitated by the Hungarian infinitive in construction (21b). The following examples have a finite and a non-finite embedded predication respectively: (21) a.

Mari lät-t-a , hogy Jänos iisz-ott. Mary see-PAsr-3s that John swim-PAsrr.3s 'Mary saw that John was swimming.'

b. Mari lät-t-a Mary see-PAsr-3s

Jänos-t

usz-ni.

John-Acc swim-iNF

'Mary saw John swim.' We assume that the presence or absence of a tense operator on the level of the embedded predication may affect the expression. The differing underlying structures of the terms with the function of Goal are (22a) and (22b), respectively: (22) a. (Past e,: [PREDICATION] (e,))^ b. ( φ e,: [PREDICATION] (e,))^, If no tense operator has been assigned, the next step in a production model of Hungarian is the following. The non-tensed embedded predication triggers some other trigger, in this case the Infinitive operator, resulting in structure (23): (23)

(e,: [Inf PREDICATION] (e,))^

In order to arrive at expression (21b), the following sequence of rules is applied: (24)

primary operator non-tensed —> embedded predication

auxiliary operator Inf

—>

Form ni

82

Casper de Groot

Another reason for distinguishing between primary and auxiliary operators is that different operations may lead to the introduction of one and the same auxiliary operator. In the example presented above, the Inf operator is introduced as a result of the non-assignment of a tense operator. In other examples some other operator may introduce the Inf operator. Operators which introduce the auxiliary Inf operator will be discussed in section 4. 3.3.

Non-finite predications in Hungarian

In Hungarian we find the following expressions of non-finite predications: (i) present participle, (ii) past participle, (iii) adverbial participle, (iv) nominalization, and (v) infinitive. In De Groot (1989) I have argued that nominalizations arise through predicate formation rules. The other four types of expression are triggered by among other things the absence of a tense operator. The choice of expressing an embedded predication as an infinitive, a present participle, a past participle, or an adverbial participle depends on (i) the status of the embedded predication as a first or a second restrictor5 of a term variable (cf. (25a-b), and (ii) the semantic function of the term, in which the predication is the first restrictor of the term variable (cf. (25c-d). The following distribution of structures yields various expressions: (25) a.

present participle (x,.: Pred„ (x,): (e,·: [PredK (x,)^ ....] (e,)))

b.

past participle (x,.: Pred„ (x,): (e,·: [PredK (x,)^ (x,)^] (e,)))

c.

adverbial participle (e,·: [PredK .... (x,) ....] ( e , ) ) ^

d.

infinitive (e,·: [PredK (x,)^ ....] (e,)))5f where, semantic function SF f Circumstance

In section 1.3, we looked at oppositions between constructions with an infinitive on the one hand, and a finite verb, a participle or a nominalization on the other. I have found no oppositions in Hungarian which express such differences as those in English and Dutch. There is a pair of constructions which resemble the above oppositions namely the pair consisting of a construction with an infinitive and a construction with a nominalization. However, there seems to be no semantic difference between the two constructions. Consider:

Auxiliary Operator Inf and constructions with an Infinitive in Hungarian (26) a.

83

Tilos dohänyoz-ni. forbidden smoke-iNF 'It is forbidden to smoke.'

b. Tilos

a

dohänyz-äs.

forbidden the smoke-NZN

'Smoking is forbidden.' The construction with the infinitive is fully productive. Despite the fact that expression (26b) is perfectly grammatical, nominalizations are seldom used in this type of construction.

4.

Constructions with an infinitive in Hungarian

In this section I will present examples of different types of construction with an infinitive in Hungarian." In describing the various constructions I will assume that the infinitive is the expression of the Inf operator and that the Inf is an auxiliary operator introduced by some other operator.7 4.1.

Non-tensed embedded predications

(Rl) Non-tensed embedded predication —> Inf Conditional) embedded predication is first restrictor (ii) term does not have the semantic function of circumstance The following examples are illustrations of embedding in the first argument of a predicate (27a), the second argument (27b), the third argument (27c), and a satellite with the function of purpose (27d): (27) a.

Könnyd Jänos-t meggyöz-ni. easy John-Acc convince-iNF 'It is easy to convince John.'

b.

P6ter usz-ni akar. Peter swim-iNF want:3s 'Peter wants to swim.'

c.

Α

tanär

elenged-t-e

the teacher allow-PAsr-3s

a

gyermek-ek-et jätsz-ani.

the child-PL-Acc

'The teacher allowed the children to play.'

play-iNF

84

Casper de Groot d. Laci elmen-t üjsäg-ot ven-ni. Laci go-PAsrr:3s newspaper-Acc buy-iNF 'Laci went out to buy a newspaper.'

The following rule, which differs from rule (Rl) may account for expressions such as (28): (R2) Non-tensed embedded predication — > Inf Conditional) embedded predication is second restrictor (ii) first restrictor is not specified (28)

Van mi-t

en-ni.

be:3s what-Acc

eat-iNF

'There is something to eat.' Examples such as (28) cannot be analysed as SoA-existentials employed by Hannay (1985), because construction (28) can be paraphrased as 'there is something to eat' and not as 'there is eating going on'. We suggest that the constructions in (28) can be analysed as existential constructions in the following manner. The predicate consists of an empty locative term. The argument position is filled by a term. The first restrictor of that term variable is not specified, i.e. the term is headless. The second restrictor is a non-finite predication. The Goal argument of the embedded predicate is bound by the term variable. ' (29)

m ^ } (X,: 0 (X,): (e,: [enK {x,)Ag (x,·: mi (x,))GJ (e,))).

It is my impression that the second condition on Rule 2 is somewhat ad hoc, and that there must be a better way to account for the occurrence of the infinitive form in this type of construction. I must first point out that in Hungarian the specification of person and number of the first argument is spelled out on the verbal predicate, for instance: (30) a.

s6täl-om walk-is Ί walk'

b. s6täl-nak walk-3P

'they walk' Note that the forms -om and -nak are referential elements which are not considered agreement markers (cf. de Groot 1989: ch.3). We may assume that in

Auxiliary Operator Inf and constructions with an Infinitive in Hungarian

85

those cases in which the first argument of a predicate is not specified for number and person, there will be no spelling out of the first argument. In those cases in which person marking on the predicate does not apply, the auxiliary Inf operator will be introduced. Compare: (31) a.

b.

s6tälK-(lp)^ s6täl-om

===>

s6t&y-W) Ag Inf s6tälK - l$)Ag sdtäl-ni

===> ===>

Since the first argument of the embedded predicate in (29) does not allow specification for person and number, the application of a rule which is based on the consideration that the Inf operator will be introduced if person marking does not apply would account for the occurrence of the infinitive in constructions such as (28). It would seem that such a rule is also needed to account for infinitives in another construction, which will be discussed in section 4.5 below. 4.2.

Future

The grammatical expression of future tense in Hungarian is an analytical expression, and consists of the auxiliary fog plus an infinitive.* (R3) may account for the introduction of the Inf operator: (R3) Future operator (32)

—>

Inf

Feri tanit-ani fog Feri teach-INF AUX.FUT.3S 'Feri will teach.'

Note that the infinitive in constructions such as (32) cannot be considered the expression of an embedded predication. 4.3.

Remote/progressive

Another example of an operator which may introduce the Inf operator is the socalled Remote/progressive operator (RemProg).9 (R4) Remote/progressive operator

—>

Inf

86

Casper de Groot

(33)

Mari enekel-ni van. Mary sing-iNF be:3s 'Mary is away singing.'

Construction (33) cannot be analysed as an existential construction, because it cannot be paraphrased as 'there is Anikö singing'. In fact, I do not think that constructions such as (33) contain an embedded predication. I suggest that this type of construction should be analysed as follows: (34)

a. b. c.

RemProg enekel,, (dlx,: Anikö (x,))^ Inf enekel K (dlx,: Anikö (x,))^ va% 6nekel r ni (dlx,: Anikö (x,))^

===> ===>

Aspect operator RemProg triggers the introduction of the Inf operator. The status of the predicate will then be that of a non-finite predicate which triggers the introduction of a copula. 4.4.

Habitual

The following type of expression is somewhat problematic. I do not know whether expressions such as (35) should be considered non-productive idiomatic expressions, or habitual-like expressions formed by some productive rule. Consider: (35)

4.5.

P6ter dolgoz-ni jär. Peter work-iNF go:3s 'Peter goes to work.' Impersonal

Hungarian allows a few cases in which the first argument of the verbal predicate is not specified for person and number. Impersonal constructions are usually formed with finite forms in the following manner: (36)

a.

Az ember nem ert-i. the man not understands 'People do not understand that.'

b.

Nem ert-ik. not understand-» 'They / people do not understand that.'

Now consider the following impersonal example with an infinitive:

Auxiliary Operator Inf and constructions with an Infinitive in Hungarian (37)

Iskolä-ban

nem tanul-ni

school-iNEssivE not

learn-iNF

ily-et,

csak az 61et-ben.

such-Acc

only

87

the life-iNEssivE

'Such things are not learnt at school, only in life.' I suggest that the following rule may also introduce the auxiliary Inf operator, (cf. example 28 above): (R5) Unspecified argument of main predication 4.6.

—>

Inf

Imperative

The imperative in Hungarian is marked by the suffix -j (cf. 16a). However, in some cases it is possible to use an infinitive construction to express the imperative:™ (R6) Imperative (38)

4.7.

—>

Inf

Fölkel-ni, h£t 6ra! get up-iNF seven hour 'Get up, it is seven o'clock!' Theme constructions

A somewhat unusual construction denoting the pragmatic structure of 'Theme predication' also uses an infinitive in Hungarian:" (R7) Theme (39)

4.8.

-- >

Inf

Hall-ni hall-om, de nem 6rt-em. hear-iNF hear-is but not understand-is 'As for hearing, I can hear it, still I cannot understand.' Derived constructions

The last type of construction in which we find an infinitive is the derived construction. There are at least two reasons to assume that the infinitives in the following examples are derived nominals: (i) the person-marking elements belong to the class of elements which combine with nouns and not with verbs, and (ii) the constituents marked by the dative case parallel the possessors in possessive constructions/2 I am assuming that one result of predicate formation is the introduction of auxiliary operators. (R8) Predicate formation

—>

Inf

88 (40)

Casper de Groot Α

fiük-nak nem volt

t h e boys-DAT

not

was

szabad vär-ni-uk. free

wait-iNF.3P

'The boys were not allowed to wait.' (41)

P6ter-nek meg kell ir-ni-a a level-et. Peter-DAT PF must write-inf-3s the letter-Acc 'Peter must write the letter.'

(42)

Α

fm-nak

t h e boy-DAT

van ki-ben is

biz-ni-a.

WIIO-INESSIVE trust-INF-3s

'The boy has someone to trust.'

5.

Discussion and conclusions

On the basis of the examples presented, I conclude that the infinitive can be the expression of either an embedded (cf. (27)) or a non-embedded predication (cf. (32-33), (38-42)). In all constructions the underlying representations of infinitive expressions designate states of affairs. We have also seen that infinitives in Hungarian may be considered inflectional (expression rule) in some cases and derivational (predicate formation) in others. If we wishes to account for the fact that in Hungarian infinitives do not arise independently or idiosyncratically in different components, it may be helpfull to view the infinitive as an expression of an auxiliary operator which can be introduced by a number of other operators. However, it is clear that the rules mentioned in section 4 are unrelated. If no explanation or motivation can be given for the fact that several apparently unrelated rules introduce one and the same auxiliary operator, then the relevance of the notion of auxiliary operator is limited. Of course, the distinction between primary and auxiliary operators is a relevant one, because it makes it possible to distinguish between those grammatical devices which have an immediate expression and those which do not. I would like to suggest a possible way of looking at the relation between the different operators which introduce the auxiliary Inf operator, based on the ideas of Myhill (1988). According to Myhill, in any clause certain morphemes can be identified as being most informationally important, while others carry less important information or none at all. In his paper, Myhill argues that clauses are pragmatically organized into clusters, each of which has one morpheme which is most informationally important, which he refers to as the nucleus, accompanied by any number of less important morphemes, which he refers to as satellites. The greater the importance of the nucleus, the more satellites there will be. Satellites show varying degrees of morphological unity with the nucleus: the closest make up

Auxiliary Operator Inf and constructions with an Infinitive in Hungarian

89

the inflectional paradigm of the nucleus; others are morphologically bound to the nucleus without being inflectional, while still others are separate words. Crucial to this discussion is the concept of categoriality, as discussed in Hopper and Thompson (1984). There are two dimensions to categoriality, specificity and anaphoricity. Highly categorial morphemes are high in specificity of reference and low in anaphoricity; thus the most categorial types of morpheme refer to a specific action or entity which has not been mentioned in the immediately preceding discourse. If this action or entity is more generic or hypothetical or if it has been mentioned so recently that it is no longer salient, it is much lower in categoriality. Examining the Hungarian data, we see that verbs marked by -ni are low in specificity of reference, because they do not refer to actual events. In some cases the infinitive even has unspecified argument positions. On the other hand, they are relatively high in anaphoricity. At least one of the argument positions of the infinitive is anaphorically bound. Continuing Myhill's argument, it can be maintained that the low verbal categoriality in Hungarian is marked by -ni. Categoriality as employed by Myhill and Hopper & Thompson, is not defined in terms of two absolute values - high and low - but rather as a continuum between high and low. Depending on a number of factors related to the quality of the action, the participants involved, and the newness, a verb takes up a position along the continuum of categoriality/3 For this reason we may find that there are gradual differences in categoriality between various applications of verbs. This seems to be the case with infinitives as well. Some expressions with an infinitive are higher in categoriality than other expressions, for instance non-actual imperfective versus non-actual perfective events. This may offer an explanation for the fact that some infinitives allow certain affixes, whereas other infinitives do not. One such example is the application of the perfective marker: perfective states of affairs are higher in specificity than imperfective states of affairs. Another example is the ability of some infinitives to combine with person-marking suffixes. These infinitives, which I consider examples of a type of nominalization (cf. (40-42)), are higher in specificity than the other infinitives in Hungarian. I realize that the ideas put forward here are as yet somewhat sketchy. However, it is my impression that the notion of categoriality will be relevant in further research on expression rules. It is in any case relevant to the study of the general principles underlying expression rules, and will contribute to our understanding of why - and not how - languages use particular bound and free morphemes.

90

Casper de Groot

NOTES

I thank Inge Genee and Harm Pinkster for their helpful comments. 1.

Hungarian does not have a distinct genitive marker. The marker -nak/-nek is used in both genitive and dative relations.

2.

Cf. Dik (1985, 1988), Mackenzie (1985), Köster & May (1982).

3.

See Myhill (1988) for the notion of specificity of reference.

4.

Cf. Dik (1989), Hengeveld (1989).

5.

In order to arrive at the correct expression, the first restrictor of terms such as (25a) and (25b) must be lexically filled.

6.

I have used the following sources: Kälmän et al. (1989), Kiss (1989), Stephanides (1980), and other literature mentioned in De Groot (1989).

7.

See De Groot (1990) for arguments which support this assumption.

8.

There are other ways to express reference to future time in Hungarian (see De Groot 1989: 9). It is not known which factors determine the choice of the different types of expression.

9.

Cf. De Groot (1989).

10. I do not know which factors determine the choice of using the imperative marked by the j or the infinitive. 11. See De Groot (1981) for a discussion of the construction. 12. I refer to De Groot (1989: 90f) for a discussion of this type of construction. 13. The more specific the reference of the verb, the more categorial it is. Thus the most categorial type of verb refers to a single action which has taken place at a single point in time, as in Then I punched him on the nose. The individual semantic properties of a verb also contribute to its categoriality. The verb walked in I walked to school has more specific reference than the verb went in / went to school, because walked specifies the means of travel. Verbs are higher in categoriality at a point in the discourse where they represent new information and are emphasized; where they represent old information, they are in the background (Myhill 1988:

2630-

REFERENCES Bartsch R. 1983. 'Over de semantiek van nominalisaties'. Glot, 6, 1-29. Dik, S.C. 1985. 'Infinitief-konstrukties met om in een Funktionele Grammatika'. Glot 8, 25-46. Dik, S.C. 1988. 'Embedded predications from a typological point of view*. Paper for the Workshop on "Typology of Languages in Europe". European Science Foundation, Rome, 7-9 January 1988. Dik, S.C. 1989. The theory of Functional Grammar. Part 1: The structure of the clause. Dordrecht: Foris.

Auxiliary Operator Inf and constructions with an Infinitive in Hungarian

91

Disterheft, D. 1980. The syntactic development of the infinitive in Indo- European. Columbus, Ohio: Slavica Publishers. Esau, H. 1973. Nominalization and Complementation in Modem German. Amsterdam & London: North-Holland. Groot, C. de. 1981. 'On Theme in Functional Grammar; an application to some constructions in spoken Hungarian'. T. Hoekstra, H. van der Hülst & Μ. Moortgat eds. Perspectives on Functional Grammar, 75-88. Dordrecht: Foris. Groot, C. de. 1989. Predicate Structure in a Functional Grammar of Hungarian. Dordrecht: Foris. Groot, C. de. 1990. 'Morphology and the typology of expression rules'. M. Hannay & E. Vester eds. Working with Functional Grammar: Descriptive and Computational applications, 187-201. Dordrecht: Foris. Hannay, M. 1985. English existentials in Functional Grammar. Dordrecht: Foris. Hengeveld, P.C. 1989. 'Layers and operators in Functional Grammar'. Linguistics, 25, 127-157. Hopper, P. & S.A. Thompson, 1984. 'The discourse basis for lexical categories in universal grammar'. Language, 60, 703-752. Kälmän, C.G., L. Kälmän, Α. Nädasdy & G. Pröszdky. 1989. Ά magyar segddigdk rendszere'. Altalänos NyeMszeti Tanulmänyok XVII, 49-103. Budapest: Akaddmiai Kiadö. Kiss, έ . Κ. 1989. 'Egy föndvi igeneves szerkezetröl'. Altalänos Nyelviszeti Tanulmänyok XVII, 153169. Budapest: Akaddmiai Kiadö. Koster, J. & R. May, 1982. 'On the constituency of infinitives'. Language 58, 116-143. Mackenzie, J.L. 1985. 'Genominaliseer'. Tekstwetenschap..., 177-198.

I'll.

Interdisciplinair

Tijdschrift

voor

Taal- &

Myhill, J. 1988. 'Categoriality and clustering'. Studies in Language 12-2, 261-297. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik. 1985. A comprehensive grammar of the English language. London & New York: Longman. Stephanides, 1980. Ά contrastive study of the English and the Hungarian infinitive'. L. Dezsö & W. Nemser Studies in English and Hungarian contrastive linguistics, 93-144. Budapest: Akaddmiai Kiadö. Vet, C. 1987. 'Infinitive incorporation'. J. van der Auwera & L. Goossens eds. Ins and Outs of the Predication, 163-177. Dordrecht: Foris.

A note on semantic functions and adpositional term predicates

Mike Hannay Department of English Free University, Amsterdam

0.

Introduction

One of the most powerful pieces of representational apparatus introduced by Simon Dik into Functional Grammar (henceforth FG) is the rule of term predicate formation (cf. Dik 1980, 1983). Together with the rule of copula support, term predicate formation allows an elegant and concise description of all kinds of so-called non-verbal predications, including adpositional predicate constructions like (1) and (2): (1) (2)

This book is for Simon This photo is of my mother when she was 12

Right from the outset, however, the rule of term predicate formation as

The research for this paper was carried out within the framework of the Free University, Amsterdam research project 'Functional language research: grammar and pragmatics' (LETT 88/10), financed by the Dutch Ministry of Education. Thanks are due to Chris Butler, Evelien Keizer and Lachlan Mackenzie for helpful comments on a draft version.

94

Mike Hannay

modified to take care of such adpositional predicates has also been a frequent bone of contention amongst people working within the FG model. The bone in point is the presence in the input to the rule of an isolated term with a semantic function, the contention being that a semantic function is a relational concept and cannot therefore be applied to a single term without there being a specified predicate or predication. This fundamental problem surrounding the rule has been the subject of varying amounts of discussion in Mackenzie & Hannay (1982: 55), Mackenzie (1983: 35), De Groot (1983: 99f, 1989: Ch.6), Hannay (1985: 50f), and Dik (1989: 174), but essentially the originally formulated rule has continued to be applied in FG descriptions. The aim of this short contribution is to propose a representation of copula constructions with adpositional predicates which succeeds in formally resolving the semantic function issue while retaining the format of a term predicate formation rule. The proposal I wish to make constitutes an application of the representation for extended term structures put forward by Mackenzie (1983), and takes the form of a term predicate formation rule which converts a satellite into a predicate. The structure of the paper is as follows. After reviewing the precise nature of the problem (section 1), I proceed to discuss an alternative analysis of adpositional predicates put forward by Weigand (1989), as well as a treatment by De Groot (1989) which attempts to provide a formal justification of the original predicate formation rule (section 2). Both are argued to be unsatisfactory solutions to the problem as a whole. I then pick up the representation of extended terms and nominal predications given in Mackenzie (1983) and show how this can form the basis for a term predicate formation rule which neatly captures the insight that adpositional term predicates constitute a derived function of satellite expressions (section 3.1). Finally I adapt the basic rule to deal with adpositional cleft constructions (section 3.2).

1. Adpositional predicates and term-predicate formation Dik (1980: Ch.4, 1983, 1989: 165f) provides an analysis of copula verbs within the framework of FG which rests on the idea that one should introduce the copula into the representations of linguistic expressions only when it is needed, rather than assuming its presence in the lexicon and then deleting it when it does not occur in the linguistic expression concerned. The consequence is a rule of copula support, which operates within the expression rule component of the grammar. The rule is formulated for English in (3):

A note on semantic functions and adpositional term predicates (3)

95

COPULA SUPPORT IN ENGLISH (fee-support) input : predicate^ (x;) (x2) ... (x„) conditions: π = any specified n-operator β f V output : η bev predicate^ (x;) (x2) ... (x„)

The expression in (4c) is then analysed as having the underlying structure of (4a), with (4b) showing the immediate result of the application of fee-support. (4a) Pres successful^, (dlx,: Kylie)^,2 (4b) Pres bev successful^ (dlx,: Kylie),,, (4c) Kylie is successful Dik (1983: 128) notes three major advantages of copula support. First of all, it promotes typological adequacy because it is not necessary to assume an underlying copula in languages that do not have one. Secondly, the introduction of the copula can be functionally motivated in that its job is to express the verbal operators that it would not be possible to express by means of non-verbal predicates. Finally, the idea of introducing the copula rather than deleting it appears to be in accordance with the fact that the absence of the copula is a general property of simplified speech registers. In accordance with this view of the copula as 'a semantically supportive element with no deep conceptual or philosophical significance' (Dik 1989: 173), cases like (5-6) are then analysed as involving a term being predicated over another term. (5)

The author of Money is Martin Amis

(6)

What PSV need is an English manager

This is what is captured by the rule of term-predicate formation, which is given in its basic version in (7): (7)

TERM-PREDICATE FORMATION (basic version) input : any term (t) output : {(t)} (χ;)φ

Next, however, we have to deal with cases like (1-2), which, following the copula support approach, will be analysed as involving adpositional term

2 I represent terms without a variable to the right of the colon, following Mackenzie (1987: 12). The same format is adopted by Dik (1989: 262), but only as a notational convention.

96

Mike Hannay

predicates. This brings us to the problem that we are directly concerned with here: Dik proposes a modification of (7) which allows for the presence of a semantic function in the input term, but no account is provided of where the semantic function comes from. The modified version is given in (8): (8)

TERM-PREDICATE FORMATION (modified version) input : any term (t) [s j where [s] indicates an optional semantic function output : {(t)[sJ} (χ;)φ

Semantic functions are assigned to terms and 'specify the roles which the referents of the terms involved play within the state of affairs designated by the predication in which these terms occur' (Dik 1989: 24). Because of the relational status that semantic functions have within a state of affairs the conclusion must be drawn that the input in (8) is in fact not formally feasible: a semantic function is assigned to a term without there being any specification of a predicate or predication which the semantic function might relate to. Moreover, if the predicate or predication is not specified, it is not possible to guarantee that an appropriate semantic function will be assigned. These objections were enough for Mackenzie and Hannay (1982: 56) to advocate a return to the idea of a lexical be for English, at least for handling the copula in identifying and locational constructions. Despite this serious inadequacy in the modified term predicate formation rule, however, it is clear that the arguments for copula support itself remain strong, and in fact the term predicate formation and copula support approach to copula constructions has become the standard analysis in FG descriptions. It must therefore be considered worth while to explore the possibilities for a solution to the semantic function problem while continuing to do justice to the principle of term predicate formation. Given the objections formulated above, it should be recognized that any adequate rule of term predicate formation must take as its input a structure which allows a clear insight into the relation between the term marked for a semantic function and some specific predicate or predication.

2. 2.1.

Two alternatives Prepositions as predicates

Weigand (1990: 148ff) offers an interesting treatment of adpositional locative constructions which might appear to obviate the whole problem addressed in section 1. The essence of Weigand's proposal is that locative prepositions should

A note on semantic functions and adpositional term predicates

97

be treated as predicates. Translating his formalism into FG representations, (9a) would thus be analysed as (9b) (prior to copula support applying), where PR stands for a prepositional predicate and Ref for the semantic function Referent. (9a) The book was in the library (9b) Past (dlx,: book)+ (dlx^: library)^ In the complex term phrase in (10a), a locative semantic function would still relate the two referents: (10a) The book in the library (10b) (dlx,: book (y,: inFÄ (dl^: library)«,^) The proposal has significant advantages for other aspects of the grammar. Most importantly, and as Weigand (1990: 96) himself points out, there is a considerable lightening of the load on locative semantic functions, which otherwise would have to be specified individually by the expression rules. Secondly, the analysis allows a neat account of prepositional phrases functioning as terms, such as the italicized expressions in (11-12) (for a similar analysis within FG see Mackenzie 1990): (11) In Gorky Park was where it all happened (12) But the real reds haven't come out from under the bed yet Nevertheless, it is doubtful whether Weigand's proposal can be extended to capture cases like (1-2) and other non-locative adpositional predicates. While such two-place prepositional predicates as in in (9b) might capture what is essentially a locational relational predication, it is clearly not the most appropriate representation for the more strongly property-assigning relation present in say (13): (13) Let it be is by the Beatles which might be paraphrased as 'Let it be is such that it was written/ performed/ sung by the Beatles'/ For such cases it appears more appropriate to assume an

Agent cannot be the relevant function in such a property-assigning predication as this (cf. also De Groot 1989: 178). Rather, something like 'accredited creator' seems to be involved. Note that out of context it appears acceptable to say her hair is by Vidal Sassoon, whereas my hair is by my sister is much less acceptable. Related to this, the entity involved tends to be interpreted as a work of art: this poem is by Baudelaire and all these novels are by up-andit took us a long time, really U U Ü j

»3

Λ

& — Exchange of information

β Η CO Η ο ä I Ο I 1

1

State of knowledge

I I I

Boundary information

1

(State of knowledge) Procedure

Farewell rituals

Figure I. The discourse structure of job interviews.

178

Martha L. Komter

The hierarchy of this structure is founded on the simultaneous operation of the different orientations, depending on their centrality in the interview. Thus the exchange of information, as central business of the interview, is embedded in additional layers of meaning, revealing additional simultaneous orientations of the participants: - the orientation to sociability, as revealed in the opening and closing stages; - the orientation to the tasks of the interviewers in organizing the interview and the selection procedure, as revealed in the explicitation of the procedure; - the orientation to the kinds of knowledge that the participants draw on to make available to one another the information that is relevant for their main tasks: the selection of the interviewers or the application of the applicant. The fact that in the great majority of the interviews these steps are taken quite explicitly suggests that this structure is easily available, not only to the observer, but also to the participants themselves. Moreover, the structure is reflexive in that it shows a continual interplay between what the participants display as their orientation at any particular moment in the interview, and how they constitute each successive step in the structure.

4.

Summary and discussion

An enumeration of the typical topics occurring at typical moments in job interviews may give a global insight into the kinds of things people talk about in job interviews: they start with some small talk, then they talk about the vacancy and the company, about the applicant, about the administrative details, and they end the interview with more small talk. People exhibit what topics are appropriate for what phases, and they interpret the topics on the basis of their positioning in the interview. However, a description of the structure of job interviews in terms of a succession of phases that can be identified on the basis of topic fails to recognize the dynamic and the hierarchic nature of the discourse structure. People move from one phase to another by way of explicit transition devices that indicate the nature of their business at that moment, and the relation of their current activities to their main official business, the exchange of information. The opening and closing rituals identify the moment as a social occasion, and the business of the participants as officially concerned with sociability; the explicitation of the procedure draws the attention to the tasks of organizing the interaction of the interview; the explicitation of the state of knowledge of the participants focusses on the official informational basis of their main business. That these steps are taken to move the participants towards and out of their main business testifies to the layered nature of their activities. The official main business does not only consist of the exchange of information, but is embedded in the additional tasks of the organization of the relevant information, the

The discourse structure of job interviews

179

organization of the speaking and listening activities of the interaction, and the organization of the job interview as a social occasion. The explicitness of the transition devices has been described in terms of the availability of the structure of the interviews. In this way the transition devices can be seen as guiding the interaction and as giving the participants a sense of what is officially going on at each successive moment in the interview. The moves that contribute to the display of the structure derive their significance from their sequential position, and, conversely, the arrival at a certain phase in the interview calls for specific moves to effectuate that transition. By taking these steps the participants continually demonstrate at what phase in the interview they are, and at what level they are operating. Thus they reflexively create and maintain the structure of the interview as the recognizable order and level of events. By means of the establishment of this discourse structure the participants provide for one another a recognizable sense of what they are doing at what moment in the talk and to what ends. I have mentioned in the introduction that the discourse structure reflects the official orientation of the participants, and that this is only a part of what goes on in the interviews. Everyone knows that important requirements for applicants are that they be likable persons and that they have the right qualifications and skills for the job. Current ideas on the objectivity and fairness of selections push the matter of the likability of the applicant into the regions of the unsayable. Moreover, applicants are unreliable sources of information on their skills and qualifications because of their interests in obtaining the job. These potential conflicts of interest are responsible for the fact that a lot of the important business in job interviews cannot be dealt with explicitly, which leads to an orientation to the 'side-effects' and to the strategic consequences of the talk. Although this paper has not been concerned with the implicit, unofficial, organization of the interviews, we may consider the significance of the explicitness with which the interviewers structure the interview, in relation to the implicitness with which much of their business is being conducted. It may be suggested that, as there is so much that cannot be said in the interviews, the interviewers make the most out of what can be said. The importance of the 'unsayables' is balanced by an 'exploitation of sayables'. This will no doubt be a result of the current demands on the professionalism, objectivity, and fairness of the interviewers. The explicitness of the ways in which the interviewers structure the interviews can be regarded as direct demonstrations of their official concern with objectivity and fairness. So the discourse structure does not only function as the recognizable order and level of events, available to the participants for the production and interpretation of their talk, but also as a constant reminder of the legitimacy and propriety of the interaction.

180

Martha L. Komter

NOTES

1. This paper is an adaptation of a chapter in my book 'Conflict and Cooperation in Job Interviews. A study of talk, tasks, and ideas' (Komter, forthcoming). 2. This study is based on a corpus of 35 authentic job interviews that have been taped and transcribed for a variety of jobs, ranging from cook to university lecturer. 3. For an analysis of the unofficial features of job interviews see Komter (forthcoming). 4. Transcription conventions: A = applicant; I = interviewer underlined = stress in the original Dutch text :: = prolongation of preceding sound .brackets .

[brackets = overla P

(.) = pause of about one second (2) = pause of about two seconds etc. ((double brackets)) = note of transcriber (single bracket) = barely audible utterance ( ) = inaudible utterance (24, p. 4) = interview nr. 24, p. 14 of transcription etc. . = falling intonation , = slightly rising intonation ? = rising intonation It should be borne in mind that the fragments are translations. They have been made so as to capture the conversational mechanisms of the original Dutch text.

REFERENCES Adelswärd, V. 1988. Styles of success. On impression management as collaborative action in job interviews. University of Linköping, Sweden. Akinasso, F.N. & C.S. Ajirotutu. 1982. Performance and ethnic style in job interviews. In: J J . Gumperz (ed.) Language and social identity. Cambridge: Cambridge University Press. Atkinson, J.M., E.C. Cuff, & J.R.E. Lee. 1978. The recommencement of a meeting as a member's accomplishment. In: J. Schenkein (ed.). Garfinkel, Η. & Η. Sacks. 1970. On formal structures of practical actions. In: J.C. McKinney & EA. Tiryakian (eds.) Theoretical Sociology. New York: Appleton Century Crofts. Goffman, E. 1971. Relations in public. Microstudies of the public order. New York: Harper & Row: Heritage, J. 1984. Garfinkel and Ethnomethodology. Cambridge: Polity Press. Komter, M.L. (forthc.) Conflict and Cooperation in Job Interviews. A study of talk, tasks, and ideas. Amsterdam: John Benjamins. Labov, W. & D. Fanshel. 1977. Therapeutic discourse: psychotherapy as conversation. New York: Academic press.

The discourse structure of job interviews

181

Polanyi, L. & R. Scha. 1984. A syntactic approach to discourse semantics. In: Proceedings of the international conference on computational linguistics. Stanford, Cal.: Stanford University Press. Sacks, H, E. Schegloff & G. Jefferson. 1974. A simplest systematics for the organization of turn-taking for conversation. Language 50, 696-735. Schenkein, J. (ed.). 1978. Studies in the organization of conversational interaction. New York: Academic Press. Raffler-Engel, W. von. 1983. Nonverbal behavior in the career interview. Amsterdam: John Benjamins. Schegloff, Ε., & H. Sacks. 1974. Opening up closings. In: R. Turner (ed.). Sudnow, D. 1972. Studies in social interaction. New York: Free Press. Turner, R. 1972. Some formal properties of therapy talk. In: D. Sudnow (ed.). Turner, R. (ed.) 1974. Ethnomethodology. Harmondsworth: Penguin.

Translation studies: developments and perspectives in teaching and research

Kitty M. van Leuven-Zwart Department of Translation Studies University of Amsterdam

0.

Introduction

Translation has often been called the oldest but one profession. Scholarly approaches to research and teaching in the field of translation cannot claim such a venerable age, though. It was only after World War II that the development of translation studies as an academic subject took off, and it was only in the later seventies that attempts were made to give a firm academic foundation to the teaching of translation and to include research as an element in this teaching. In the Netherlands this new approach is realised since 1982, with the foundation of the Department of Translation Studies at the University of Amsterdam. It was Simon Dik who stood at the cradle of this department, and for years the epithet 'godfather of the department' stuck to him. This is why I want to take a closer look here at the prospects of teaching and research in translation studies in general, and at the future development of our department in particular. The indelible link between past and future prompted me to trace the origin of translation studies as an academic discipline, which is dealt with in section 1. In section 2 a brief look is taken at the teaching of translation as an academic subject, and at the roots of our department. Section 3 sketches ways

184

Kitty Μ. van Leuven-Zwart

in which translation studies can develop into a mature and independent discipline. Section 4 discusses the consequences such a development in the field would have for teaching and research at universities, and section 5 contains a summary and conclusion.

1. Research and Translation Studies: Developments It is generally agreed that translation studies emerged as an academic subject sometime in the late forties. Until then, there certainly had been a long and fruitful tradition of translation as well as of reflection on, and study of, translation: translators, theologians, philosophers and philologists had tried to find answers ίο the questions posed by translation. But in the years after World War Π a renewed interest in translation emerged, this time from disciplines such as linguistics, logic, mathematics, and information theory; an interest partly induced by the high expectations people had in those days of the possibilities of machine translation. Not surprisingly, translation researchers of this period aimed at the highest possible degree of objectivity, taking the research methods of the natural sciences as their model. This limited approach was unable to deal with anything not directly pertaining to the phenomenon of language. Matters such as the situation in which language is used or the intention of a speaker or writer were not taken into account: translation research of this period was a matter of grammar only. As a consequence, literary translation was hardly a legitimate subject of study, the general feeling being that with literary translation such vague and unscientific matters as intuition and creativity play an essential part, precluding a systematic, objective approach. Accordingly, that period's research into literary translation characteristically lacked any attempt at scholarly objectivity/ Thus, in post-war translation studies two distinct approaches emerged, a linguistic and a literary one, split by a deep rift. Linguistic translation research at that time was dominated mainly by the socalled Leipziger übersetzungswissenschaftliche Schule,2 to which the names of people such as Neubert, Jaeger and Kade are attached. Rather than exploring translation issues, however, they dealt with contrastive grammar. Their research took place on the level of, what De Saussure terms, 'langue', whereas translation is typically a 'parole' phenomenon, to use another famous Saussurean term. Central to translation is, indeed, the use that is made of the

1

See Apel (1983: 8-9).

2

See Koller (1978: 78-79) and Snell-Hornby (1988: 14).

Translation Studies: developments and perspectives in teaching and research 185 language system; translation is not a matter of interpreting language systems but of interpreting language utterances. With this insight, which has been termed 'one of the main achievements of the theory of translation in the last decades' (Toury 1980:103), the second stage of linguistic translation research has been reached, in which Eugene Nida, American Bible translator and researcher, figures prominently. Nida's point of departure, too, is linguistics, and more especially semantics, but he does include aspects that were excluded by the Leipziger Schule. This is why he looks for links with various other disciplines, such as the philosophy of language, communication studies, stylistics, cultural antropology, psychology and sociology. Many of the ideas underlying linguistic translation research in progress in the United States and Europe, especially in Germany, can still be traced back to Nida. Curiously, most linguistically oriented works in the field approach translation from a didactic, prescriptive angle. In spite of promising titles, often featuring the word 'theory', it is not the translation phenomenon that is being explored in these studies, but rules and prescriptions for the production of a good translation that are given. This prescriptive approach is not limited to linguistic studies, either. The same tendency can be seen in literary translation research, where most 'theories' turn out to be manuals for the literary translator.3 It is only in the course of the seventies that the prescriptive approach to both linguistic and literary translation research was called into question. The first sounds of protest came from literary researchers such as Holmes, Toury, Van den Broeck and Lefevere. They claim that translation research should be undertaken not from a prescriptive, but from a descriptive angle/ This new look-out is of great importance, especially for the research into literary translation. Questions as to what constitutes a good translation and what is the best way to translate, matters of endless dispute in the past, have become obsolete and irrelevant. A new field of research is opened up in which as late as 1988 it is still claimed that 'nearly everything' is yet to be done (Hermans 1988: 20). Unfortunately, this revolution did not culminate in the emergence of one independent discipline harbouring both linguistic and literary translation research. On the contrary; the line dividing the two is sharper than ever. Hermans, for instance, states that linguistics is relevant in research dealing with the translation of so-called unmarked texts but is unsuited for research in the field of literary translation:

3

Examples: Mounin (1963), Catford (1965), Levy (1969), Koller (1972), Nida & Taber (1974).

4

See Holmes, Lambert & van den Broeck (1978).

186

Kitty Μ. van Leuven-Zwart 'Linguistics has undoubtedly benefitted our understanding of translation as far as the treatment of unmarked, non-literary texts is concerned. But as it proved too restricted in scope to be of much use to literary studies generally - witness the frantic attempts in recent years to construct a text linguistics - and unable to deal with the manifold complexities of literary works, it became obvious that it could not serve as a proper basis for the study of literary translations either' (Hermans 1985: 10).

Speaking from the other side of the fence, Wolfram Wilss declares the translation of 'unmarked' texts to be the only subject of 'Übersetzungswissenschaft', claiming that in literary translation the elusive quality of the 'freie Spiel sprachgestalterischer Kräfte' (Wilss 1977: 181) plays too great a part. The rift between the two approaches is deepened further by methodological disputes, with Toury and his disciples, for example, insisting that translation studies be purely descriptive and the linguistic Übersetzungswissenschaft regarding translation studies as an applied science which consequently can only be purely prescriptive.

2. Teaching and Translation Studies: Developments In this section I will look briefly at the background of translation studies as an academic subject, in particular in relation to my own department. This department is part of the Faculty of Arts of the University of Amsterdam. It was founded in 1982, the result of a merger between the Translation Institute, a training institute for professional translators founded in 1964, and the university course of Translation Studies, which only dealt with the theory of translation. That Simon Dik has played a major part in the foundation of the Department of Translation Studies has already been mentioned. As a chairman of the Translation Institute's Board of Foundation he pleaded for years for the inclusion of the Institute in the Faculty of Arts. This was to serve two objectives: translation training would have a firm academic base, and there would be a breeding ground for future researchers in translation studies. Dik was appointed chairman of the Committee for Integration and under his chairmanship integration became a fact. Until the appointment of the first professor in translation studies in 1984 he also acted as representative of the department. This development in the field of translation teaching in the Netherlands is not an isolated occurence. In the course of the seventies all over Europe the same tendency manifests itself. The call for well-trained professional translators is heard more and more, and there are new developments in research. This leads

Translation Studies: developments and perspectives in teaching and research 187 to an increasing demand that practical translation training be based on theoretical insights. Throughout Europe then, translation training courses are being incorporated in university departments, and everywhere attempts are made to turn these vocational courses into academic ones. Nowhere in Europe does this change take place without a hiccup, and Amsterdam is no exception. Though all faculty members were happy to come under the umbrella of the Faculty of Arts, it soon became clear that the move involved far-reaching and sometimes unforeseen changes. Most teachers of the new department were professional translators with a vocational rather than a theoretical background, and with hardly any research experience in the field of translation studies. All of a sudden, these teachers were faced with the problem of having to link theory with practice; practical translation training was to be given an academic foundation, and research was to be an essential component of the course. In the years following the merger, several steps were taken to forge this link between theory and practice. One such step has been to organize seminars on subjects connected with teachers' current research. But especially in the practical part of the course, the link between theory and practice falls far short of the ideal, and, again, the situation in Amsterdam is not exceptional. Too often the teaching of translation skills is basically assumed to imply trial-anderror methods, and consequently to involve a great deal of individual coaching. The effectiveness of such methods has never been proved, and there is a pressing need to look for a different approach: higher education throughout Europe is hit by cuts, and when time is money and money is short, such laborious methods are too expensive. Translation studies as an academic course would benefit from an integration of theory and practice, both in research and in teaching, in terms of efficiency as well as effectiveness.

3.

Research and Translation Studies: Perspectives

A link between theory and practice of research and teaching in translation studies can only be fully established if there is clarity about the subject and the aims of such teaching and research. For translation research this clarity is achieved with the publication of James S Holmes' seminal article, 'The Name and Nature of Translation Studies' (Holmes 1988). In his article Holmes makes use of a distinction Francis Bacon made between two kinds of research, i.e. research which is 'of light' and research which is 'of use'. 'Of light' here is the pure or theoretical research in translation studies, which has as its objective to describe and explain the phenomena of translating

188

Kitty Μ. van Leuven-Zwart

and translation(s).5 Holmes distinguishes three areas in this field.6 In the first area the processes taking place in the translator's mind during translation constitute the object of the research. In the second area the focus is on the characteristics of existing translations. Research here may take the form of a comparison of translation and original. Aim of this type of research is to establish translators' norms and strategies in different periods of time and in different cultures.7 In the third area the role played by translations in the recipient culture is the object of investigation. Holmes sees all these three areas as relevant to both literary and non-literary translation. Translation research which is 'of use' is called 'applied translation studies' in this article. The aim of this kind of research is to develop means and methods which are of direct use to the translator. Again, Holmes distinguishes a number of areas here. In one area the reseacher's concern is with translation dictionaries, translation grammars and terminology banks, whether or not computer-linked. Two other important areas are translation criticism and translation training. In translation criticism the aim is to develop methods with which translations can be evaluated with more justification than has been usual up till now. Research in the area of translator training looks for ways in which teaching can be made more systematic and more effective. As in the case of pure translation research, this may apply to both literary and non-literary translation. In my opinion, 'The Name and Nature of Translation Studies' has prepared the ground on which the discipline can grow to maturity. Holmes creates the conditions here which are needed to dissolve the three controversies that have been dominating the field up to this very day. The first controversy is a matter of the rift and rivalry between linguistic and literary research. Holmes' view is that both belong to the field of translation studies. At the same time, Holmes underlines with this view the interdisciplinary character of the discipline. Not only linguistic and literary research, but also, say, text linguistic, psycholinguistic and semiotic research are legitimate subjects of research in translation studies. s In fact, Holmes distinguishes two branches of pure translation studies: 'As a field of pure research ... translation studies thus has two main objectives: (1) to describe the phenomena of translating and translation(s) as they manifest themselves in the world of our experience, and (2) to establish general principles by means of which these phenomena can be explained and predicted. The two branches of pure translation studies concerning themselves with these objectives can be designated descriptive translation studies or translation description and theoretical translation studies or translation theory(Holmes 1988: 71). I combine the two branches as they cannot be considered in isolation. 6 In Holmes (1988) these three areas belong to the branch of descriptive translation studies. In the branch of theoretical translation studies six areas are distinguished, which, in fact, cannot be considered in isolation from the three descriptive areas (see note 5). 7 See Toury (1980) and van Leuven-Zwart (1984), in which methods for the comparison and description of translations are developed.

Translation Studies: developments and perspectives in teaching and research 189 The second controversy is the above mentioned dispute about the role of translation studies: is it to be prescriptive or descriptive. Holmes accomodates both views. The descriptive approach is of its nature connected with pure translation research, whereas the prescriptive approach is best suited to the demands of applied translation research. In Holmes' view these two research areas can only come to full bloom if they feed each other continually; lacking that, they will turn sterile.® This view has completely dissolved the question whether translation research should be prescriptive or descriptive. The third controversy is between the theory and the practice of translation, and their respective practitioners. Holmes contends that researchers, both in pure and in applied translation studies, have neglected to link their activities to the practice of translation, thus widening the gap between theory and practice, and leading translators to either ignore or belittle the efforts of theoreticians in the field.9 The danger is, Holmes claims, that eventually the absence of a strong link between theory and practice will dam up research. Translation, of course, will go on as usual. It has managed to do without translation studies research for centuries and will not suddenly come to a halt for the lack of it. But this is no denial of the fact that the practice of translation can benefit from research; in section 5. I shall demonstrate and illustrate this. There is reason enough for attempts from that side of the field, too, to bridge the gap between theory and practice. Academic translation courses in general, and the Department of Translation Studies in particular, can play a key role in the development outlined by Holmes - the development of translation studies into a mature discipline. In their teaching and research such courses have a permanent link with both the theory and the practice of translation. Thus, the conditions for the strongest possible link between theory and practice, both in research and in teaching, are all present. So, how far has its realisation come about? In research several steps have been taken to prepare the ground for such an integration. Research in the department takes place both in the field of linguistics and in that of literature,

s . '...descriptive, theoretical and applied translation studies have been presented as three fairly distinct branches of the entire discipline, and the order of presentation might be taken to suggest that their import for one another is unidirectional, translation description supplying the basic data upon which translation theory is to be built, and the two of them providing the scholarly findings which are to be put to use in applied translation studies. In reality, of course, the relation is a dialectical one, with each of the three branches supplying materials for the other two, and making use of the findings which they in turn provide it... In view of this dialectical relationship, it follows that...attention to all three branches is required if the discipline is to grow and flourish.' (Holmes 1988: 78-79). 9 'Wie in een gezelschap vertalers het woord vertaalwetenschap laat vallen kan slechts hoongelach verwachten' (Langeveld 1986:11) ('...the mere mention of the discipline [of translation studies] among translators is enough to provoke jeers and laughter'(van Leuven-Zwart 1988: 701).

190

Kitty Μ. van Leuven-Zwart

both in pure translation studies and in the applied branch of the discipline. In addition, the department regularly produces a variety of translations in the fields of literature, culture and science. What is needed now are stronger connections across the various activities and across the various fields. One such connection could be made by increasingly incorporating the production of translations in the research in progress. As most university translation courses have at least some members of staff who are also acclaimed translators, this would be quite a natural undertaking. Such links across the fields must be made if the role of the department, and of university translation courses in general, is to take its proper shape.

4. Teaching and Translation Studies: Perspectives The connection between the various fields of research, as well as between research and practice, can be the basis for the link between theory and practice in the teaching of translation. Clarity about the objectives of teaching in this field is a prerequisite, though. In my opinion, a university translation course should produce competent translation scholars who have a firm grasp of their subject. This means that they have the knowledge and the skills to develop into future independent researchers or full-fledged translators, or, in exceptional cases, into both. Which way they will go, depends on their interests and abilities. Thus, students who aim at a career in the practice of translation - the majority - must be able to present themselves on the job market as competent translators. This means that they know what they are capable of and that they know what they are doing when they are translating. These two requirements must be reflected in the teaching. To become aware what they are capable of, students must be exposed to not only, say, literary or legal texts, but to the greatest possible variety of texts. In other words, students are to end up with translation skills, not with proficiency in the translation of one or two specialised subjects. To become aware of what they are doing - the second requirement in translation teaching - students must be confronted with the consequences of the decisions they take in the process of translation. One way of doing this is to bring students in touch with the different norms and strategies translators apply. This involves the comparison of existing translations with their originals. Thus students are made aware of the choices taken and the consequences such choices have for the text as a whole, since certain choices may affect, in varying degrees, such aspects as the function of a text, the portrayal of characters, and the reflection of actions and events. To portray, say, a character who is aware of the relativity of things as dithering and bewildered, would constitute a

Translation Studies: developments and perspectives in teaching and research 191 definite shift of emphasis; again, to translate an article on drugs in the USA from a Dutch or an American point of view would have a very different impact. The requirement that students be made aware of what they are doing when they are translating, takes us back to the connection between theory and practice. There is a direct link between the two here, as the consequences of the choices taken in translation and the norms and strategies dictating these choices is one of the subjects of pure research. Of course this is not the only area which shows a direct and useful link between theory and practice. Applied research, too, affords many possibilities to link theory with practice. My emphasis on the link between pure research and translation teaching has been deliberate, though. So often it has been the theoretical branch of translation studies which has been regarded as irrelevant to the practice of translation. It is my conviction that its relevance is irrefutable. When students are taught to see what other translators have done, they may gain insight in what is going on when one translates. When students are taught to see in a systematic and academically sound way from a variety of angles what other translators have done, they are given the instruments to make the right choices according to the demands of each text, when they are translating on their own. By learning to make choices, being aware of the choices they make, and anticipating the consequences of such choices, students can develop into translators who know what they are doing. Systematic interweaving of theory and practice would clearly enhance the effectiveness of translation training. A welcome extra would be the increased efficiency one may expect to result from such a link. Surely, if students know what they are doing, their translations will improve and such improvements will result in time saved, as a relatively bad translation takes more time to correct than a relatively good one.

5.

Conclusion

To summarise, translation studies will only develop into an independent and mature discipline, if and when cross-fertilisation takes place between the different areas. A hopeful sign in this context is the appearance of the first international periodical for translation studies, Target, edited by representatives of different schools of thought. The development of translation studies into a mature discipline is of the utmost importance for the future of university translation courses. But this development will not materialise of its own accord, and translation courses may play a substantial part here by continually seeking to establish the greatest possible connection between the theory and practice of translation, both in their research and in their teaching. It is my opinion that this should be a matter of

192

Kitty Μ. van Leuven-Zwart

course, as university translation courses exist only by virtue of such a link. They can only produce qualified and capable translators and researchers when they have managed to embrace and interlink the theory and practice of translation.

REFERENCES

Apel, F. 1983. Literarische Übersetzung. Stuttgart: J.B.Metzler. Broeck, R. van den (red.) 1988. Literatuur van elders. Over het vertaten en de Studie van vertaalde literatuur in het Nederlands. Leuven: Acco. Catford, J.C. 1965. A Linguistic Theory of Translation. An essay in Applied Linguistics. London: Oxford. Hermans, T. 1985. The Manipulation of Literature. Studies in Literary Translation. London & Sidney: Croom Helm. Hermans, T. 1988. Van 'Hebban olla vogala* tot Ernst van Altena: literaire vertaling en Nederlandse literatuurgeschiedenis. In: Van den Broeck (1988). Holmes, J.S 1988. TranslatedI Papers on Literary Translation and Translation Studies. Amsterdam: Rodopi. Holmes, J.S, J. Lambert and R. van den Broeck (eds.) 1978. Literature and Translation. New Perspectives in Literary Studies. Leuven: Acco. Koller, W. 1972. Grundprobleme der Übersetzungstheorie. Bern & München: Fincke. Koller, W. 1987. Einführung in die Übersetzungswissenschaft. Heidelberg/Wiesbaden: Quelle & Meyer. Langeveld, A. 1986. Vertalen wat er stoat. Amsterdam: De Arbeiderspers. Leuven-Zwart, K.M. van 1984. Vertaling en origineel. Dordrecht: Foris. Leuven-Zwart, K.M. van 1988. Vertalen en vertaalwetenschap: tweespalt of eenheid. De Gids, honderdeenenvijftigste jaargang nr. 9, 701-709. Levy, J. 1969. Die literarische Übersetzung. Theorie einer Kunstgattung. Frankfurt a/M: Athenaeum. Mounin, G. 1963. Les probUmes thioriques de la traduction. Paris: Gallimard. Nida, Ε A. & CA. Taber 1974. The Theory and Practice of Translation. Leiden: Brill. Snell-Hornby, M. 1988. Translation Studies. An Integrated Approach. Amsterdam: John Benjamins. Target International Journal of Translation Studies. Amsterdam/Philadelphia: John Benjamins. Toury, G. 1980°. Communication in Translated Texts, a Semiotic Approach. In: Wilss 1980. Toury, G. 19806. In Search of a Theory of Translation. Tel Aviv: The Porter Institute for Poetics and Semiotics.

Translation Studies: developments and perspectives in teaching and research 193 Wilss, W. 1977. Übersetzungswissenschaft. Probleme und Methoden. Stuttgart: Klett. Wiks, W. (Hrsg.) 1980. Semiotik und Übersetzen. Tübingen: Gunter Narr.

The development of cases and adpositions in Latin

Harm Pinkster Department of Latin University of Amsterdam

'Speech has no need of those additional words. So it is, it seems to me, with the speech of the Romans, which now is used by nearly all men, for it has eliminated all prepositions except for a few1 (Plutarchus 24 Quaest. Plat. X.1010D) *We have much more (prepositions) than the Greeks' (Prise. 14,9)J

0.

Introduction

It is a well-known fact that in Latin case plays an important role in marking the functional relations of nominal constituents of a clause, much more so than in the Romance daughter languages. Conversely, in the Romance languages adpositions (pre- and postpositions) are used with a higher frequency. Various explanations can be found for the change that took place, the most prominent among which is of a phonological nature: it is often said that the attrition of the inflexional endings and the ensuing merging of cases brought about the increase in the use of adpositions. In this paper I will examine the adpositional system and try to show that as early as the archaic and classical period the adpositions

1

Plutarchus lived from ca. 45 - 120. Priscianus lived in the sixth century. I owe the reference to Prof. Dick Schenkeveld.

196

Harm Pinkster

had certain characteristics of their own. From this we might conclude that adpositions are not simply analytical alternatives for case endings which may take over the functions of the case system if that system collapses.

1. From Latin to Romance: a preliminary survey Latin had a mixed marking system in which both adpositions and 'bare' cases (i.e. cases not governed by adpositions) were used to indicate syntactic and semantic functions of nominal constituents. The relative frequency of both types of marking varies from author to author and from text type to text type. In one of Plautus' comedies (Amphitruo) adpositions constitute 3.75% of the vocabulary, whereas Caesar in his de bello Gallico has 8.61% (compare Pinkster 1990: 42).2 There is also a distributional difference between the two types of marking. Bare cases are used for first, second and third arguments, as well as for satellites and attributes.3 Adpositions are used for third arguments (ex. 1), satellites (ex. 2) and attributes (ex. 3), as is illustrated by the following instances of the preposition de: (1)

qui ... de veneficiis accusabant whoTOm of poisoning^, they accused (*who accused < some people > of poisoning', Cie. S.Rosc. 90)

(2)

de vehiculo dicebat from wagon^ he spoke ('he spoke from a wagon', Nep. Timol. 4,2)

(3)

de collegio quis... adfuit? of college^ w h o ^ was present (Svho was present of the college?', Cie. Dom. 117)

Adpositions are also used for marking second arguments, though not very frequently. In the case of satellites, adpositions are almost as frequent as bare cases, as we will see below (cf. Pinkster 1990: 65). There are a number of

2

Elcock (1975: 34) observes that in Cicero's letters we find 'the increased reliance upon prepositions which is one of the decisive characteristics of Vulgar Latin'. As the percentages given in the text suggest, many factors may influence the frequency of adpositions. In Plautus the low percentage certainly is also due to the high frequency of adjectives. In Caesar the relative frequency of movement verbs may explain the relative frequency of directional and locative satellites and hence of directional and locative prepositions. 3 'Argument' and 'satellite' are used in the sense of Dik (1978), that is for obligatory and optional constituents with respect to the predicate, respectively.

The development of cases and adpositions in Latin

197

expressions where adpositions and bare cases seem to be mere alternatives, for example with the verb liberare ('to release from'). I will come back to this below. However, there are also adpositional expressions for which no bare case expression exists. An example is the preposition sine ("without'). There is no bare case expression for 'withoutness'.'' Conversely, adpositions are excluded in the case of first arguments. In the Romance languages case marking plays a less important role. In fact, in most Romance languages - Rumanian being an exception - it disappeared almost completely. This can be shown easily with the help of the following two tables. 'Bare cases' in these tables means 'nominal constituents not governed by an adposition'. Table 1: Proportion of bare cases and adpositions in Nepos Milt. 1-2 (Latin and modern translations)

Latin Italian5 French

bare case 85% 40 50

adposition 15% 60 50

Table 2: Distribution of bare cases and adpositions in Sail. Hist, Or. Lep. 1-7 (Latin and Italian translation)

Latin bare case adposition Italian bare case adposition

1 arg 25 0 18 0

2 arg 13 0 16 2

3 arg 6 2 1 6

sat 16 10 0 21

attr 15 0 0 21

total 75 12 35 50

The two tables show that the increase of adpositions is most significant in the domain of attributes and third arguments, but there are also interesting instances of the use of adpositions for marking second arguments. This increase can be illustrated by the use of Romance successors of the Latin preposition ad ('to') for third arguments with verbs of giving and communication, and by the spread of successors of Latin de to all types of attributive noun phrases (cf. ex. 3). The intrusion of successors of ad with two-place verbs is most 4 Adpositions without a bare case alternative can be found in particular in the 'secondary layer' of adpositions (Lehmann 1982: 94). 5 The Italian translation is by V. Paladini (Bari, Adriatica Editrice, 1957). The French translation is the Budd edition.

198

Harm Pinkster

remarkable: Fr. nuire ä, It. nuocere a correspond to Latin nocere, which governs a dative. There is almost no sign for this development in our sources.0 With these verbs the 'irregularity' of Latin (dative instead of accusative) continued in French and Italian. These languages adopted the ad-expression that had come into use instead of the Latin dative with verbs of giving and communication. In the domain of satellites there is an increase in the use of adpositions. However, adpositions were strongly represented in classical Latin as well.

2.

Causes and mechanisms

The change that took place in Latin may be explained as being due to a reduction of the case system, as an expansion of the adpositional system (or as a mix of both). 2.1.

Reduction of the case system

As I stated in the Introduction, the increase of adpositions is often regarded as a compensation for the collapse of the case system, which, in turn, is often ascribed to phonological and morphological changes. I start with the phonological changes. The absence of final consonants in nominal forms can be documented from the archaic inscriptions onwards. It is found with the nominative -s ending of the nominative singular of the second declension (ex. 4, dated approx. 250 BC), but the most remarkable instances are those of absence of final -m in a wide range of occurrences (ex. 5, dated somewhat later), which brought Väänänen (1981: 66) to his neat formulation: 'le -m final 6tait caduc d£s l'6poque archaique'. (4)

Cornelio Cornelius^ (CIL I2,8)

6 However, there are a few interesting expressions in the Mulomedicina Chironis, especially if one compares them to the text of Vegetius, for which theMul. Ch. was a source. Vegetius' Latin is much more 'classical' than his source text. Compare haec res et ad febricantes prosunt (Mul. Ch. 115,29) with quae potio etiam febricitantibus (dat.) prodest (Veg. Mulom. 206,9). See Grevander (1926: 103-105).

The development of cases and adpositions in Latin (5)

hone this^

oino one^

ploirume most^

cosentiont agree

199

duonoro good^

optumo fuise viro best^ have been m a n ^ ('Most people agree that this man has been the best of the good men', CIL I2, 9) Instances where final -s is lacking are not abundant in inscriptions and other texts (on the latter cf. Carlton 1973) even very late, although in poetry syllables ending in -s are peculiar in some way (cf. Allen 1978 : 36f.). On the other hand the evidence for the loss or absence of final -m is overwhelming and there are even early ancient testimonies to confirm the absence of the -m in the pronunciation of words ending in -m (cf. Allen 1978: 30ff. with references and the collection of ancient testimonies in Kramer 1976: 58-63). In the classical period a trace seems to have remained in the form of nasalisation and lengthening of the preceding vowel. The notation of -m in classical authors and documents of the classical period therefore can be regarded as an attempt to express the still existing phonetic difference between nasalized and plain final vowels in spelling. Later, probably not before 150 AD (Beckmann 1963: 180ff.), the presence of -m in writing is due to conservative spelling. The loss of final -m had serious consequences for the distinction of the ablative and accusative singular forms, sometimes also the dative singular forms. Accusative mensam and ablative mensa merged. This means that in the singular the amount of homonymy was much larger than would appear from our literary tradition. At first sight this may have been a fatal stroke to the case system. However, the formal distinction of accusative and ablative continued to exist in the plural. The occurrence of plural forms is not that much less than corresponding singular forms. Therefore, it cannot be easily assumed that the changes in the singular system dragged the plural forms along, witness table 3: Table 3: Relative frequency of Accusative and Ablative forms of nouns in Caesar Gal.

acc. abl.

sg. 19% 22%

pi. 12.5% 12%

Furthermore, whatever happened to the accusative and ablative, there remained several forms which stayed outside the development mentioned above. Finally, the merging of ablative and accusative sg. did not take place with the same

200

Harm Pinkster

speed in all declensions. In the second declension the changes of accusative (-om > -urn > -o) and ablative (-o:d > -o: > -o) must have taken more time than in the consonant declension (Beckmann 1963: 180 ff.). The conclusion to be drawn from this section, then, is that it is implausible that phonological developments alone, or predominantly, had left the case system with a problem to solve, which the system then solved by handing over some of its tasks to the adpositional system. I now come to the morphological development. In Latin, different inflexional endings are used for different declensions. There is not only an evolution with respect to the differentiation of cases, but also with respect to the distinctness of the declension classes. In fact, the statistically infrequent u- and e- stems (as in fructus 'fruit' and dies 'day') tended to become absorbed into the more frequently used classes, e.g. fructus into the o-stems (as in hortus 'garden'). Inscriptional evidence shows that the declensional paradigms were not as straightforward as we are taught in school. Examples like collegas instead of collegae (nominative plural form of COLLEGA 'colleague') and patris instead of patribus (dative plural form of PATER 'father') (Gaeng 1984: 19, 81 resp.) show that inflexional endings were taken over from one declensional class to the other. The precise cause-effect relationship between the morphological and phonological developments is difficult to assess.7 The very existence of morphological variation, however, confirms the insufficiency of a purely phonological approach. 2.2.

Expansion of the adpositional system

Several explanations have been suggested for the increase of adpositions which are based on certain inherent properties or qualities of adpositions. I will now turn to these. One quite general explanation - often tacitly understood and sometimes overtly expressed - that has been suggested for the development is based on the assumption of some sort of intrinsic 'drift' towards analyticity, which is sometimes linked to the assumption of a desire for expressiveness, in particular in the so-called 'Volkssprache'.* In such a view the evolution of the Romance languages is just a continuation of a similar development from Indo-European to Latin. However, given the fact that this putative drift does not manifest itself in all Indo-European languages and even in all Romance languages does not

7 Carstairs (1987) gives some theoretical background for the relationship between morphological readjustments and phonological changes. 8

Cf. Svennung (1936: 362): 'Die Volkssprache liebt ja die expressive Ausdrucksweise'.

The development of cases and adpositions in Latin

201

manifest itself in the same way and to the same extent, drift as an overall cause does not help very much. A more specific mechanism mentioned in this context is the gradual replacement of a bare case expression by an adpositional expression which though not synonymous is very close in meaning, or at least in implication. Petersmann (1977: 90-1) is one of the scholars who point to a 'Verschiebung der Vorstellung' as an important factor for the spread of prepositional expressions in Latin. The spread of ad ('to') is a case in point. Ad as a substitute for the bare dative is well-known both in the case of purpose expressions (so-called dativus finalis, cf. Szantyr 1965: 86,93,220) and, most prominently, in the case of third arguments with verbs of communication and giving. One of the early examples referred to in this context is (6): (6)

uxor... ad vi rum nuptum datur wife^ to m a n ^ to many is given ('a wife is given to a man in marriage', PL St. 140)

The example is interesting because the manuscript partly read ad virum, the adpositional expression, and partly viro, the bare dative. Petersmann, who later tended to date the substitution of ad for the dative rather late, defends the adpositional expression as lectio difficilior. Parallels cited in support of (6) are (7) and (9): (7)

haec me modo ad mortem dedit thisMm meacc just to deathacc gave ('she has handed me over to be killed', PL Am. 809)

(8)

ibo ad medicum atque ibi me toxico morti dabo I will go to doctoracc and there meecc poisonous^ death^ I will give ('I'll go to a doctor and end it all with poison', PL Mer. 472)

(9)

facere ut det nomen ad Molas coloniam to make that he gives nameacc to Molae^ colony^ ('to make him sign up as an emigrant to Millcolonia', PL Pseud. 1100)

(10) ei ego urbi Gripo indam nomen t h a t ^ \ m m town^ Gripus^ I will give name«, ( Ί will call that town Gripusburg', PL Rud. 934)

202

Harm Pinkster

The difference between the ad expression in (7) and the dative expression in (8) is that in (7) an element of destination is involved,9 in contrast with (8) where the handing over in its literal meaning is at stake. The difference between (9) and (10) is brought out well by the Loeb translation. The same holds for verbs of communication, as in ex. (11): (11) quae ad patrem vis nuntiari which^ to father^ you want to report ('the message he's to carry to your father', PI. Capt. 360) Although muitiare + dat. is quite normal, there is a difference in meaning between the two expressions. The ad expression conveys the idea of transportation towards someone, whereas the dative would mean 'to communicate to'. Similarly, dicere ('to say') + ad is not the same as dicere + dative. Apart from a semantic difference there may be a structural difference as well. In the former construction (with ad) nuntiare and dicere may be regarded as two-place verbs with an optional ad satellite. In the construction with the dative they are three-place verbs. The historical development might therefore be described as the grammaticalisation of the preference of the 'change of place' aspect over the 'coming into the possession of aspect. Another mechanism is the spread of an adpositional expression from one context where it is quite appropriate to another where it is at first sight out of place. The working of this mechanism has been illustrated by Beckmann (1963) in a detailed study on the development of instrumental expressions, for example with ad ('to'), de ('from'), ex ('out o f ) in ('in'). I take some of his examples of ex. The earliest example of ex in an instrumental expression, disputed as it is, is nonetheless interesting to illustrate part of the development: (12) si forte ex nare sagaci sensit (canis) if by chance animals^,, from noseaW sagacious^, has noticed dognom ('if by chance a dog smells wild animals with his sagacious nose', Enn. Ann. 341V(=333Sk)i0 Particularly frequent in Celsus and other medical authors (cf. Englund 1935: 117-9) are expressions of ex with nouns indicating liquid material (like aqua Svater') in connection with verbs like eluere ('to wash clean') and unguere ('to annoint'). The plain ablative case is equally possible in these contexts. The ex

9

10

Cf. ad camificem in PI. Capt. 1019. For reservations about the validity of this example see Skutsch a.l.

The development of cases and adpositions in Latin

203

expression can be understood as derived from either a partitive or a source meaning ('to clean with some water' or 'to wash by taking some water'). From this lexically restricted context it spread to other contexts, a late example of which is (13): (13) ex digitis exploramus from fingers^, we investigate (Sve search with our fingers', Mart. Cap. 9,968) The same mechanism (spread to other domains) can be demonstrated in the case of other Instrument/Manner expressions. A closer examination of the use of the prepositions cum ("with') and per (through') in archaic and classical Latin may also shed some light on the function of adpositions in marking nominal constituents. The difference between Instrument and Manner expressions is not always easy to establish. It is to some extent related to the lexical meanings of the nouns and verbs involved and often the presence of qualifying adjectives is a clue to the semantic function of the NP as a whole (cf. Vester 1983). By instrumental expressions I will here understand objects, actions and methods wilfully employed by some person in order to reach a specific goal (cf. Beckmann 1963). In Latin the ablative case is a very common means of marking Instrument/Manner constituents. However, from the earliest texts onwards the adpositions cum and per are used as well. Among the instances of cum in the Thesaurus Linguae Latinae (s.v. 1363-1370), expressions with concrete nouns as their prepositional complement (type gladius 'sword') appear to be rare. I discuss a few of the examples: (14) in solo tracta cum melle oblinito bene in crustaW tractaocc with honey^, cover well ('cover the tracta on the crust thickly with honey", Cato R. 77) (15) ista cum lingua... possis culos lingere thataWwith tongue^, you can anusecc lick (Svith a tongue like that you might lick the anus', Catul. 98,3) (16) (olea) melior ... quae digitis nudis ... oliveMm better„om which„om fingers^, bare^,

quam than

ilia quae cum digitabulis (legitur) that mm whichnom with gloveseW is picked ('the olive is better which is picked with bare hands that that one which is picked with gloves (Varro RR 1,55,1)

204

Harm Pinkster

With the verb oblinere of (14) ('to smear') the thing used is normally expressed by the bare ablative/ 7 This notable exception may be explained, however, because of the contrast with another similar recipe a few lines further on, which has sine melle ("without honey')/ 2 Something similar may be the case with ex. (16). The contrast is between nudis and cum. However, this example is not a straightforward instance of an instrumental expression either. Clothes can hardly ever be viewed as instruments. In this example the action of picking is performed with the hands, wearing gloves. It would be better to regard the cumphrase as a Praedicativum (cf. Pinkster 1990:146). Ex. (15) has evoked various explanations in the commentaries on Catullus. Quinn, for example, observes that 'cum with the instrumental is an archaism', which cannot be true. Kroll regards the cum phrase as denoting a property of the subject: 'You with your...', which, in my opinion, is correct. Most of the examples in the Thesaurus Linguae Latinae receive similar explanations.75 Examples with a concrete noun, in fact, come late, for example (17): (17) caede caudam cum tabula aliqua ponderosa cut tail^ with plank^, some^, heavy^ ('cut the tail with some heavy plank'. Veg. Mulom. 1,27,2) We see, then, that the use of the preposition cum in its 'instrumental' meaning obeys quite narrow selection restrictions. The preposition is allowed with concrete nouns if the concept of instrumentality is to be emphasised or if an Instrument interpretation is not quite obvious. The latter explanation might also hold for Manner satellites: if the noun phrase has an adjective the interpretation normally seems to be clear enough. If it is lacking, potential ambiguity is avoided by using cum.

11 I treat oblinere as a two-place verb, although there is some reason to regard it as a three-place verb on account of the existence of dat. acc. alternatives, just as with circumdare ('to surround') 12

13

It is even conceivable that tracta cum melle is a noun phrase with a prepositional attribute.

I mention a few of the more embarrassing ones. (a) cistella... cum crepundiis quibuscum (quibus Fleckeisen) filiam inveni meam ('a litle casket of toys which enabled me to find my daughter', PI. Rud. 1363). The Loeb translation makes clear that the casket and its content were not real instruments in the action of finding the daughter. This makes the presence of cum no less remarkable. (b) tibiae ..., cum quibus in funere canitur (Paul Fest. 33,1). The cum phrase indicates the accompaniment, not the instrument, cf. Plin. N H 7,204 cum tibiis canere voce. (c) ut etiam cum acopo ... articuli perfricentur (Cels. 4.31). Edd. read tum and Loeb translates: 'that also then the joints may be rubbed with an anodyne salve'. (d) quem cum manu tangeret (Ps. Quint, decl. exc. Monac. 358,21 L) I do not fully understand the text. (e) per quod utaris cum eo utaris. (Tert. Resurr. 7, p. 36,18). Some of the manuscripts omit cum eo utaris.

The development of cases and adpositions in Latin

205

Something similar seems to apply to per. In Plautus this preposition is used in Instrument/Manner expressions7'' like (18) and (19): (18) urbe capta per dolum town aw capture^, by guile^ ('after the town had been captured by guile', PI. Ba. 1070) (19) haec vobis dixi per iocum theseacc y o u ^ I said through f u n ^ ('this I said to you in fun', PL Poe. 542-2) Bare case expressions parallel to the prepositional expressions in (18) and (19) do exist/ 5 albeit with a much lower frequency. It seems that the adpositional expressions are highly idiomatic. Lodge mentions a few examples of instrumental per with concrete nouns in his dictionary (1924), but they are all questionable. I discuss one example given by Lodge: (20) strictim adtonsurum ... an per pectinem closely going to trim,,,.,, or through comb"" ('is it going to be a close crop or just a trim', PL Capt.268) This is a particularly interesting example, since the prepositional phrase is coordinated with the adverb strictim, which is normally taken as a manner adverb. However, the comb is not the proper instrument with which the cutting is performed. The bare ablative pectine seems to be impossible. As far as I can see, the instrumental use of per is limited in the case of concrete nouns in classical prose. This is certainly true for Cicero70 and Caesar. According to Muhr (1971: 95) Sallust prefers the use of per with abstract nouns, which sometimes could not be used in the bare ablative. The selection of the per expression is mainly due to either Sallust's aiming at variation in expression or the avoidance of possibly ambiguous ablatives in the same clause (see Muhr 1971: 97). Later on, Tacitus, who shares Sallust's preference for varied

14 Per is often used with animate beings to indicate the intermediate agency, as in an ab imprudente aut per imprudentem sit interfectus (Quint. Deel. 248, p. 18,17). It also is used with a causal meaning ('due to'). It is not always easy to distinguish pure instrumental/manner use from these uses. 15 For example expugnare dolis ('to capture by guiles') in PI. Mi. 1157 and ioco dixisti fyou said this in fun') in PI. Am. 964, taking up per iocum in the preceding line, and being in contrast with serio ac vera. 16 In Cicero expressions like per litteras (as in Cie. Alt. 12,14,3 feci ut ipse me per litteras consolarer) are closest to the instrumental use of per. per expressions are quite frequent in the Rhet.ad Her. in definitions like concessio est per quam nobis ignosci postulamus (2,23).

206

Harm Pinkster

expressions (Ku£era 1882: 16), has a few examples which look similar to bare ablative expressions77 indicating Instrument. In technical prose adpositions in general, and per in particular, seem to be more frequent. Englund (1935: 1203) discusses various examples of per in Celsus. A clear example of a medical instrument expression with per is (21): (21)

adducta per hamulum vena praeciditur drawnnom by hookacc v e i n ^ is cut off ('the vein drawn forward by the hook is cut off, Cels. 7,13,3)

Two factors are mentioned by Englund which may explain the relative frequency of per. The first is the overall preference for adpositional expressions with pronouns/* The second is the avoidance of more than one ablative in a clause. I would like to add a third factor myself, namely the fact that in texts of this kind ('instructions') the agentless passive is used quite frequently (cf. ex. (21)). In such a context the ablative might be interpreted either as the Instrument used in performing the action or as the Force responsible for causing the action. The existence of such factors which may influence the choice of the adpositional expression shows that we are still far away from late examples like (22): (22)

si loca sancta per idola polluissent if placesoccholyacc through statuesac(. they had polluted ('If they had polluted the holy places by statues', Hier. Ep. 58,3)19

Generalising from these observations on cum and per, the division of tasks between bare cases and adpositional expressions with the function Instrument/Manner seems to be the following: concrete instruments have to be marked by the bare ablative, whereas with abstract nouns (including verbal nouns) both bare case expressions and adpositional expressions are possible, with a preference for adpositional expressions, especially in certain idioms. There is no real free variation. Selection restrictions may be ignored if the instrumental relationship is emphasised or if ambiguity might arise.

17 For example resumpta per arma dominatione ('recovering their throne by force of arms', Tac. Hist. 5,8,3) and servos per tormenta interrogari placuit ('it was resolved to question the slaves under torture', Tac. A. 2,30,3 - cf. tormentis in 3,67,3). Parallells for the latter example can be found in legal texts.

Cf. Löfstedt (1942: 274,297f.), who assumes that the preference of adpositions is due to the small size of pronouns. However, the avoidance of ambiguity may have been a factor as well. 19

Cf. Goelzer (1884: 333)

The development of cases and adpositions in Latin 3.

207

Conclusion

In the final part of the preceding section I tried to show that adpositions are used with a much higher frequency with abstract than with concrete nominal expressions. This association between adpositions and abstract nouns has been observed in other fields as well. Thdoret (1982) has examined verbs which allow alternative bare case and adpositional expressions and has noted a similar association between adpositions and abstract nouns for Cicero. The alternative expressions that occur with the verb liberare can be shown not to be interchangeable either (see Pinkster 1990: 267). According to Pitkaränta (1978: 48ff.) late Latin instrumental de is primarily found with abstract nouns in literary texts. Concrete nouns seem to be restricted to Vulgar' registers. In all these cases we are dealing with nominal expressions closely connected to the predicate (as arguments) or to the nuclear predication (Instrument/Manner satellites, that is level 1 satellites in the sense of Dik (1989: 192)). We might hypothesise that the frequency of adpositions with abstract nouns derives from the fact that the semantic relationship between these nouns and the predicate (or nuclear predication) is less predictable than in the case of concrete nouns. In the case of satellites that have no close relation to the predicate or nuclear predication (level 2 or higher satellites) adpositional expressions were the normal thing. As I have observed before, there is no bare case expression for Svithoutness', in any period of Latin that we can study. We must conclude, therefore, that in archaic and classical Latin there was a clear, basically semantic, motivation for the use of adpositions, which, as such, is independent of phonological or morphological properties of the nominal expressions involved. If it is correct that adpositions fulfilled a specific function in the structure of clauses, being more frequent the less they are related to the predicate or nuclear predication, and if, moreover, they were preferred in certain lexical contexts (abstract nouns, contrastive situations), then this might have some consequence for our idea of the nature and the chronology of the decrease of bare case and the increase of adpositional expressions. Starting with chronology, scholars differ greatly in their opinion about the relative chronology of the replacement of bare case expressions by adpositional expressions. Salonius (1920: 96) holds that adpositional expressions existed under the surface of our literary tradition where they 'unter der Eisdecke der normalisierenden klassischen Prosa ein kümmerliches Dasein fristeten', an opinion which is shared by Mohrmann's (1947). Speaking about the prepositions used instead of the instrumental ablative she observes: 'Cette tendance remonte sans doute ä des temps anciens' (= ßtudes III, p. 220). Differences as exist between the vulgar Itala text of the Bible and its less vulgar counterpart, the Vulgata, and between the vulgar Mulomedicina Chironis and its less vulgar

208

Harm Pinkster

adaptor Vegetius seem to point to some awareness among the more educated people that in certain expression types prepositions ought to be avoided. On the other hand, however, given the massive use of adpositions for all types of semantic relations it is very remarkable that the selection restrictions with respect to concrete nouns, including real instruments, were respected so consistently. This is an additional consideration for a rather late date of the replacement of bare case expressions by adpositional ones.20' The persistence of selection restrictions furthermore suggests the picture of a gradual spread of the adpositional expression type, which found its last victim, so to speak, in expressions which were quite understandable on the basis of their lexical meaning. With the verbs of giving and communication, for example, the verbs themselves are excellent indicators of the meaning relations involved. In the case of nouns meaning 'instrument' their possible relationship with respect to the action denoted by the verb usually is not very problematic. The existence of lexical restrictions on the use of adpositions may be regarded as an indication that the use of adpositions is not the result of phonological attrition or at least solely a result of this attrition. The attrition, on the other hand, may have been facilitated by the independent spread of the use of adpositions.

REFERENCES

Allen, W.S. 19782 Vox Latina. Cambridge: University Press. Beckmann, GA. 1963. Die Nachfolgekonstruktionen des instrumentalen Ablativs im Spätlatein und im Französischen (Beihefte ZRPh. 106). Tübingen: Niemeyer. Carlton, Ch.M. 1973. A Linguistic Analysis of Late Latin Documents Composed in Ravenna between AD 445-700. The Hague: Mouton. Carstairs, A. 1987. Allomorphy in Inflexion. London: Croom Helm. Dik, S.C. 1978. Functional Grammar. Amsterdam: North-Holland. Dik, S.C. 1989. The Theory of Functional Grammar. Dordrecht: Foris. Elcock, W.D. 1975. The Romance Languages. London: Faber & Faber. Englund, J. 1935. Kasussyntax hos Α. Cornelius Celsus. Diss. Göteborg.

20 Gaudemet (1979: 58-59) places the development of ad far too early (I owe the reference to dr Klaas Worp). I quote two clear examples from the TLL: (a) non quasi ad extraneam sed quasi ad uxorem fecit (seil, donationem) (Ulp. Dig. 24,1,32,27) (b) munera dantes ad servientes ibidem (Itin. Hier. Anton. 30).

The development of cases and adpositions in Latin

209

Gaeng, PA. 1984. Collapse and reorganization of the Latin Nominal Flection as Reflected in Epigraphic Sources. Potomac: Scripta Humanistica. Gaudemet, J. 19792. La formation du droit siculier et du droit de l'tglise auxIVe et Ve siicles. Paris. Goelzer, H. 1884. La Latinitd de St. Jirome. Paris. Grevander, S. 1926. Untersuchungen zur Sprache der Mulomedicina Chironis. Lund: Gleerup. Kramer, J. 1976. Literarische Quellen zurAussprache des Vulgärlateins. Meisenheim am Glan: Hain. Ku&ra, E. 1882. Über die taciteische Inconcinnität. Programm Olmütz. Lodge, G. 1924. Lexicon Plautinum. Leipzig (= Hildesheim, Olms, 1971) Lehmann, C. 1982. Thoughts on Grammaticalization I. AKUP 48. Löfstedt, Ε. 19422. Syntactica I. Lund: Gleerup. Mohrmann, C. 1947. 'Quelques observations linguistiques ä propos de la nouvelle version latine du psautier', Vig.Chr. 1, 114-28;168-82 (= Etudes sur le latin des chrdtiens III, 197-225). Muhr, O. 1971. Die Präposition per bei Sallust. Diss. Graz. Pinkster, Η. 1990. Latin Syntax and Semantics. London: Routledge. Pitkäranta, R. 1978. Studien zum Latein des Victor Vitensis. Helsinki: Societas Scientiarum Fennica. Salonius, A.H. 1920. Vitae Patrum. Lund: Gleerup. Svennung, J. 1936. Untersuchungen zu Palladius und zur lateinischen Fach- und Volkssprache. Uppsala: Almquist & Wiksells. Thdoret, M. 1982. Les discours de Ciciron. La concurrence du tour casuel et du tourpripositionnel. Montr6al: U.P. Väänänen, V. 1981^ Introduction au latin vulgaire. Paris: Klincksieck. Vester, E. 1983. Instrument and Manner Expressions in Latin. Assen: Van Gorcum.

Adverbs in Papuan languages. Or: "Where have all the adverbs gone?"

Ger P. Reesink State University, Leiden

0.

Introduction

Traditionally, adverbs are considered as one of the main parts-of-speech, although it is conceded that the term is often applied to several sets of words in a language. Nevertheless, Schachter (1985: 20) attempts to give a functional definition: "Adverbs are modifiers of verbs, adjectives or other adverbs; in general: modifiers of words other than nouns, modifiers of complete sentences or verb phrases." It can be rather confusing if the function of 'modifier' is equated with the category 'adverb', as in Givön (1984: 77), who says that "many adverbs are full sentential constructions." Therefore we need to follow Greenbaum (1969: 2) and more recently McCawley (1988:583) in distinguishing between the notion 'adverb' and the more general notion of adverbial expression. My aim in this paper is to investigate the part-of-speech adverb in Papuan languages. Although I will disregard adverbial phrases and clauses, I will signal some other word classes that are used to convey adverbial notions. In contrast to the large open class of adverbs in languages such as English (Quirk et al. 1985), the descriptive grammars of Papuan languages either give a relatively small class of adverbs (Bruce 1984: 87 for Alamblak; Davies 1981: 49 for Kobon; Reesink 1987: 69 for Usan; Roberts 1987: 73 for Amele; de Vries 1989: 49 for Wambon and Kombai; Bunn 1974 for Golin, etc.) or do not

212

Ger P. Reesink

mention the wordclass at all (Anceaux 1965 for Nimboran, Voorhoeve 1965 for Asmat). Haiman (1980: 235) only lists a few locative adverbs for Hua. Firstly, I want to consider the definition of adverbs in various Papuan languages. To what extent does syntactic function play a role in the definition? Is it possible to use morphological criteria as well? Various grammars also include a category 'particle', which seems a rather vague waste-basket category of elements that resemble adverbs with regard to their function. Secondly, given the apparent dearth of adverbs the question rises what other devices Papuan languages have to express adverbial meanings? After some illustration of other devices I will discuss whether the elements identified as adverbs express operators or satellites (cf. Dik et al. 1990). Finally, at what level do they operate? Ernst (1984) is one of many linguists who distinguish sentence and verb phrase adverbs. Dik's Functional Grammar (1989) differentiates four levels. Can adverbs in Papuan languages be identified with specific levels? And is this one or the only parameter that can be used for subclassification? Given the heterogeneity of Papuan languages (see e.g. Foley 1986), it would seem rather presumptuous to base the topic "adverbs in Papuan languages" on such a limited number of descriptions. Since the languages under consideration are sufficiently diverse, while showing some striking similarities, I will nevertheless venture some general observations that could serve as a basis for further research.

1.

Definition of adverbs

Most of the grammars mentioned list time words and locatives (sometimes referred to by other terms) as separate categories apart from adverbs and/or particles. Adverbs (and particles) are described as free uninflected forms that modify the predicate or clause. (Note: Since I am referring to a number of descriptions, it would be rather cumbersome to quote the definitions verbatim, because various kinds of terminology have been used.) 1.1.

Temporal adverbs

It would seem that at least in some languages time words may be considered as adverbs. Take Kobon rol 'tomorrow/yesterday' as in (1).

213

Adverbs in Papuan languages (1)

Kobon: Rol yad yesterday I

aip Jepi aip abö g-öl with J. with draw bow-SIM

mid-ai-lo ribö giyan de-DUR-RPID river below 'Yesterday Jepi and I were lying in wait down at the river' Rol is listed as a general time expression, together with, among others, haino 'later, after', porpor 'always' (Davies 1981: 140). These latter two words figure as adverbs on p.49. In (2) rol conforms to Davies' stipulation that simple adverbs "must immediately precede the predicate". (2)

Kobon: Yad rol ar-nab-in Is tomorrow go-fut-ls Ί will go tomorrow'

Combining these bits of information I would suggest that temporal adverbs in Kobon express temporal satellites and that they are relatively free as regards their position. This characterization is not unlike the behaviour of the Usan temporal adverb barim 'one day removed' (i.e. *yesterday' or 'tomorrow') in (3). (3)

Usan:

Barim uru uyouboun tomorrow dance we will sway 'Tomorrow we will dance the "singsing"'

Rather typically for Papuan languages, the time reference may be done by an adverb, an adverbial phrase, such as qen ger-t 'at one time' or even a clause with a medial verb, as in (4) (4)

Usan:

Worom is-a ginam weib di-arei sun descend-3s.DS place arrive come up-3s.FP 'When the sun went down, he arrived at the village'

For Alamblak, Bruce (1984: 85) mentions a formal similarity between time words and adverbs and suggests that this may reflect similar clausal functions of time words and adverbs. The former obligatorily take the postposition -none 'in, on, at', while the latter may also occur as free uninflected forms. Compare the behaviour of the time word dbha 'morning' and the modal adverb bumbri 'quickly', as Bruce calls it (p.87), both as Head of the general setting phrase in (5) and (6):

214

Ger P. Reesink

(5)

Alamblak:

dbha-nane mett tonhemetr nuam morning-in woman she.fried.for.him sago 'In the morning the woman fried sago for him'

(6)

Alamblak:

bumbri-t-nane pailatr fakrekntabmart hurry-3sF-Relator pilot switched.he.it 'The pilot switched over in a hurry'

Apparently, temporal adverbs express temporal satellites, they are monomorphemic, they cannot co-occur with other phrasal elements such as nouns, adjectives, demonstratives and the like, and they cannot be defined by a particular syntactic position. 1.2.

Spatial adverbs

With regard to locative words, it is not clear whether these include adverbs at all. Haiman (1980: 235) states that some Hua adverbs are clearly locative forms of nouns, e.g. auvaroga 'near, close by' (from auva 'body'), and that others, which cannot be so analyzed, can all be used as adjectives or nouns. Adverbial meanings of location or direction are generally expressed by nouns or deictics or combinations of these word classes. Deictics are also used as determiners in nominal constituents. It would seem that the following Hua and Alamblak examples do not represent the word class 'adverb': (7)

Hua:

Kiko'-ka-roga-mo ri-ka u-o place-your-ALLAT-PT take-2SG go-IMP 'Take (them) to your place'

(8)

Hua:

Ma-roga-mo ri-bai-ro-hi-e here-ALL-PT take-PROG-HAB-BEN-fin.A 'He kept taking them over here'

(9)

Alamblak:

Yima-m tekthemb-t tone-me-m-t man-3PL riverbank-3sF run-R.PST-3pl-3sF 'Men ran (along) the riverbank'

These examples agree with the facts from Usan (Reesink 1987: 71), for which I argued that most locational adverbs (note that I should have said "adverbials") are spatial deictics or adjectives. This is not to deny that Usan seems to have a few true spatial adverbs: e.g. qataben 'on the way', daan 'away from water'. These words are always uninflected, they do not co-occur with nouns, adjectives or deictics in a nominal constituent, and they do not seem to be intimately

Adverbs in Papuan languages

215

linked to certain verbs (cf. next section). Nevertheless they express a spatial satellite, as (10) and (11) illustrate. (10) Usan:

qanambam eng daan big wogub... goanna the away.from.water put.SS cease.SS 'having put the goanna away from the water...'

(11) Usan: in Will nob ombur daan iraminei we W. with two away.from.waterwe.went.up 'Will and I went up away from the water' 1.3.

Manner adverbs

Even though Roberts (1987: 73-75) gives a list of various subclasses of adverbs, he also states explicitly (p. 158) that adverbs are not formally distinguished from nouns or verbs. He gives (12) as example to show that all pre-verbal elements can freely move around, including the "manner adverb" mahuc 'quickly. (12) Amele: Naus uqa cum ene qa sab mahuc ute-i-an N. he yesterday here dog food quickly 3S-3S-PAST 'Yesterday Naus quickly gave the dog food here' With (13) he shows that mahuc is not really an adverb because it functions as a nominal. (13) Amele: Mahuc nij-i-a quick lie-3S-tod.p 'The quick (one) is here' The small classes of adverbs identified as such all seem to express some kind of manner. Davies does not subclassify them for Kobon. He stipulates that these simple adverbs must immediately precede the predicate (but see above). Bruce provides some subclassification for Alamblak and adds the category 'particle', which is obligatorily uninflected, whereas the adverbs may be uninflected. For Usan I stated that manner is generally expressed by nouns or verbs and then recognized a class of 'modal particles'. De Vries (1989: 48,49) makes similar comments about Kombai and also includes a class of adverbial particles. The discussion is gradually leading into my next topic: what other devices do Papuan languages have to express adverbial meanings? Before I proceed with that I want to attempt a general characterization of 'adverb' that may be applicable cross-linguistically to Papuan languages. Adverbs in Papuan

216

Ger P. Reesink

languages are words that 1) are optionally or obligatorily uninflected; 2) modify words other than nouns, complete sentences or verb phrases (cf. Schachter's definition given earlier); 3) cannot be reduced to adjectives, deictics, nouns or verbs. Forms listed as adverbs in various languages are suspiciously alike as far as their English glosses are concerned. I will give representative samples in order to refer to some of them in the rest of this paper: (14) Golin:

wen para kirara

Very, thoroughly, well' 'enough' 'completely· (Bunn 1974: 63)

Alamblak:

wompam bi bumbri nhofje masat inji

'again' 'already' 'quickly' 'slowly, carefully' Very, much' 'thus' (Bruce 1984: 87)

Kobon:

abe agamij hado dip iharin kayan si nono

'completely' 'unobserved' 'completely' 'quickly' 'only, just' 'again' 'illicitly' 'truthfully' (Davies 1981: 49)

Amele:

cebit gauc haun heje mahuc mele bahic

'slowly' •uselessly' 'again' 'illicitly' 'quickly' 'tml/ Ver/

(Roberts 1987: 74)

217

Adverbs in Papuan languages Usan:

mesenau ar bo urim qand sig mom gor

'slowly' 'uselessly' 'again' 'illicitly' 'quickly* Very' 'completely" (Reesink 1987: 72-76) 'also'

It would seem that English translations of such adverbs are rough approximations: 1) mom means 'completely", 'already*, 'altogether'; 2) Kobon has dip, igid, lei all with the gloss 'quickly*, as well as three morphemes meaning 'completely'; 3) forms translated by 'only, just, purpose- or use-lessly' often are translated in Tok Pisin as noting, which carries all these and other meanings. In spite of Amele example (13) and Roberts' claim that Amele does not have true adverbs, I would expect that the words haun 'again', bahic Very and perhaps others as well would not be different from their counterparts in other Papuan languages.

2. Other devices Roberts, as mentioned, claims that all "adverbs" in Amele are either noun or adjective. That languages short on adverbs make use of adjectives (i.e. without regular morphological derivation, which changes word class as in English, French, etc) is not exceptional (cf. Schachter 1985: 22). Usan is another Papuan language that makes adverbial use of adjectives: (15) Usan:

wo apand qura he true he speaks 'He speaks truthfully'

The word apand 'true' is an adjective in nominal phrases such as munon apand eng 'the true man'. Secondly, as Schachter (1985: 22) says, "in some languages the meaning equivalent of certain adverbs is expressed by verbs". In Papuan languages there is the use of serial verb constructions in which the modifying verb, expressing the "adverbial notion of manner", immediately precedes the main predicate, as observed by de Vries (1989: 48) and Reesink (1987: 73):

218

Ger P. Reesink

(16) Wambon: Jakhov-e matet-mo they-CONN good-SUPP.SS 'Did they travel well?' (17) Usan:

ka-lembo go-3PL.PAST

Ne mi musub me qoyam baian you thing do well.SS not knowledge you take 'You are not learning anything well'

This mechanism is either explicitly or implicitly shown for many other languages, for example, Suena (Wilson 1974: 30), Golin (Bunn 1974: 72), Yessan-Mayo (Foreman 1974: 94). Thirdly, as Foley (1986: 117ff) pointed out, many languages employ "combinations of adjunct nominale and generic verbs". I should qualify this statement by saying that the adjuncts need not be nominals, e.g. Kobon uses what Davies (1981: 201,202) calls incorporation of adjective, adverb, particle as well, and that the genericness of the verb in these constructions is a matter of degree. Thus gi 'do' in Kobon, or hu 'do' in Hua (see Haiman 1980: 117), are quite generic: (18) Kobon: kub gi 'to grow' = adjective incorporation big do abö gi 'to tread on' = particle incorporation completely do (19) Hua:

soko hu 'be good' good do

The verbs pu 'rise' in Asmat and inab 'lie down' in Usan are rather less generic. They have a more specific meaning and do not allow as wide a divergence of adjuncts as the verb 'do': (20) Asmat: binip pu-m 'to reflect' thought rise-CAUS (21) Usan:

mungam inab sleep lie

'to sleep'

Fourthly, some languages allow even more complex predicate formations, the nature of which is not quite clear. That is to say, a number of morphemes may be compounded into one verb which is then affixed for the regular personnumber categories. Whether this process is simply a verb serialization or noun

219

Adverbs in Papuan languages

(or other category) incorporation, or a combination, I cannot evaluate at this moment. The semantic complexities of such constructions in Alamblak (22) or Asmat (23) (Voorhoeve 1965: 46) are hardly less than those of highly synthetic Eskimo or Caucasian languages (cf. Schachter 1985: 23). (22) Alamblak: Yenr nur-nheh - hetfaschild cry-feignedly- from place to placehase me r CONT.AT- REM.PAST- 3SM Ά child continually cried feignedly from place to place' (23) Asmat:

sisimkakami- mtiw shove-RED- cause-pushed.down- RED- cause-lie.down 'to wash ashore many objects'

The Asmat complex is amenable to regular verbal affixation. The last three ways of specifying additional features of the State of Affairs can be analyzed in a Functional Grammar approach as different types of predicate formation.

3.

Operators or satellites

Although the samples of adverbs given in (14) are rather similar semantically, it remains hazardous to suggest an analysis that claims validity for all these languages, because there are hardly any examples with those adverbs in context. There is not much doubt, however, with regard to temporal adverbs and - if any exist - spatial adverbs. They clearly express satellite terms, because they are lexical means to provide additional information concerning the state of affairs. Some of the manner adverbs of (14), it seems to me, are neither operator nor satellite, but should be seen as incorporated in the predicate. In other words, they are the result of a process of predicate formation. All Papuan languages have elements glossed as 'illicitly', see (14), which together with some verb of 'taking' express 'stealing': (24) Usan:

Wo yonou narau he my betelnut 'He stole my betelnut'

urim ba qorei illicitly take he went

220

Ger P. Reesink

(25) Kobon: Nipe wafiib si ud ar-öp he stringbag illicitly take go-perf.3s 'He stole the stringbag' Usan has a form ende 'thus', which marks adverbial manner clauses, as in (26), or it is used, just as Alamblak inji, as a pro-form substituting for an adverb or other adverbial constructions, as in (27). (26) Usan:

an munon qei mi qand-qand you man some thing quickly-RED oboumon gas ende ebetemei you.do like thus I.did Ί acted like some of you men (who) do things rashly'

(27) Usan:

wo ende me abau he thus not do.NOM 'He did not like that'

Now, it is not possible to let ende refer to urim only: (28) Usan:

*wo narau urim baurei. Ye ende me baiau he betelnut illicitly he.took I thus not take.NOM 'He stole betelnut. I did not'

If the manner of taking is to be negated, the pro-form ende would have to refer to the whole predicate: ye ende ue Ί thus not'. Consider now the following examples from Usan. I can only suggest that corresponding items in other languages function similarly. ar : 'just, purposeless, 'nating" (29) a

b

taui qamb ar ab g-un pay back.SS say.SS just call.SS see-lPL.SUBJ 'let's just try to get a compensation' ar ai tap degen ir nanaui-mboun just ground path long ascend.SS short winded-1PL.FUT 'We'll just get exhausted going up a long path'

in Papuan languages

221

bo : 'again, in turn' a

ye bo ar sabar qemer-ib-am I again just on top say-SG.FUT-lSG 'I'll just say it again anew*

b

ye ar bo sabar qemeribam

c

ye ar sabar bo qemeribam

d

simbiri wab inab s. wipe.SS lie.SS ne and

igoumon e you(PL)are this

bo an mi ger gab ba again you thing one see.SS take

weri-s-ar them-give-PL.IMP 'Since you have been flouting (my advice), again you look for something and give it to them'

gor : 'also' a

Ne and

gor an also you

mi igo thing be.SS

meim-t qamb for what say.SS

me bigoumon? not you put down.PR 'Also, why aren't you putting down anything?' b

Qamami unum qei gor qamb igour in-law.uncle name some also say.SS they.are 'They are also saying some (other) victory cries' (Qamami unum = idiom) mom : 'completely, enough, already'

a

qanam umo mom apami ba... beginning down completely demolish.SS take 'taking it down completely to the ground...'

222

Ger P. Reesink b

uru ginam dance place

ago eng outside the

mom completely

ai ground

wat mom ue maisi wogub... break compl. not clear.SS cease.SS 'after he had broken the ground of the dance ground outside and completely finished clearing the place...' It would seem to me that ar 'just, only, etc' in (29), bo 'again' in (30), gor 'also' in (31) and mom 'completely, already in (32) do not represent satellite terms for the following reasons: a. Irrespective of the number of satellites (which is rather restricted in Usan, as in other Papuan languages), one or more of these adverbs can always be added; b. They have a preference for the position immediately preceding the predicate, although this is not an absolute requirement; c. They indicate some kind of modality, i.e. reflect an evaluative attitude of the speaker; d. They are easily repeated in consecutive clauses, which most "true satellites" do not allow (cf. Reesink 1987: 75,76); e. These adverbs cannot be replaced by full NPs or clauses that would express a similar satellite relation to the predicate. This contrasts with the deictics, locative words and temporals discussed in section 1 above. If they do not express satellites, they would be operators within a FG framework. Dik (1989: 186) characterizes operators as follows: - they provide additional specification of the SoA, - do not change the predicate frame, - are not expressed by lexical, but by grammatical means, - have the regularity of application and expression which is commonly associated with inflectional rather than with derivational processes. The adverbs we are considering do not meet each of these characteristics. They do not change the predicate frame, and do provide additional specification, but they do not resemble inflectional processes and are lexical expressions. I would be happier with a third category, but for want of a satisfying one I will take them to be operators.

223

Adverbs in Papuan languages 4. Level of operation

Κ we apply the test of negation, it would appear that most of the operators function at the predicate level (cf. Dik et al 1990: 55ff). Consider: (33) Ye mom me is-inei I completely not descend-ls.Neg.F

#

oru qas is-ibam momentarily only descend-ls.FUT Ί won't go down completely. I will just go for a time' (i.e. I won't stay) (34) wo bo me man is-au he again not garden descend-Nom 'He did not go down to the garden again' (35) se qas me yar-au now only not come-Nom 'It did not come just now' (i.e. it came long ago) The three adverbs (i.e. "operators") mom 'completely*, bo 'again' and qas 'only* fall under the scope of the negative. This contrasts with the predication satellite se-t in (36), which is outside the scope of negation. (36) se-t qur-t qas me qamarau now-at money-for only not say-Nom 'At first they did not talk only about money' The term qur-t 'for money' also seems to be a satellite at the predication level: (37) wuri qur-t ete quraur:"..." they money-for thus they.say 'They say about money thus:"...'" But without the limiting adverb qas in (36) negation would affect this satellite first: (38) se-t qur-t me qamarau now-at money-for not say.NOM 'At first they did not talk about money'

224

Ger P. Reesink

These facts suggest that qas is an adverb that functions as a limiting operator at the predicate level. Gor 'also' seems to operate on a higher level, since it cannot be affected by negation. This is true for gor in both positions of (31), i.e. whether it precedes all the terms including Subject, as in (31)a, or whether it is positioned immediately preceding the predicate, as in (31)b. In the former position gor parallels English 'also' as an additive conjunct, in the latter position as an additive adjunct, which Greenbaum (1969: 49,50) identified as homonyms. In my opinion, it would be better to recognize one form gor 'also' as additive adverb which may function at different levels: gor in (36)a is outside the interrogative illocution; in (36)b it is part of the predication. In the case of bo, the position also makes a difference. Notice that in (30)a (30)c bo may take various positions, but as long as it appears between the subject term and the predicate it will be affected by negation, as in (34). When bo precedes the full predication, as in (30)d, it functions at the predication (or proposition?) level, because it cannot be affected by negation, nor by the other sentential modes, interrogative or imperative (see for syntactic similarities between interrogative, imperative and negation Reesink 1987 chapter 7). Thus, in (30)d bo 'again' does not fall under the scope of the imperative of the final verb. The different instances of bo parallel Greenbaum's (1969: 47) differentiation of English 'again' as additive conjunct and adjunct. Again, I would suggest that we are dealing with the same adverb which can operate on different levels, as McCawley argued for English adverbs such as 'quickly', 'carefully', etc.: (39)

Marvin carefully sliced all the bagels carefully

Bo 'again, in turn' may even occur within a noun phrase to modify a quantifier, which function is shared by its counterparts in e.g. English and Dutch: (40) wai sarag-au bo ger igub ... animal rustle-Nom again one hear.SS '(he) heard yet another animal's rustling noise and...' The additive adverb bo then can operate on a whole predication or proposition (30)d, a predicate (30)a-c, or a quantifier within a NP (40). The limiting adverb qas 'only' operates on a satellite (36), on a nuclear argument (41) or a predicate in a verb serialization (42). (41) Tari qas aib burab head only big become.SS 'The head only grows and...'

Adverbs in Papuan languages

225

(42) Ye qobur qas quraum I cut.along.the.grain only Lspeak Ί speak only straightforward' The degree adverb sig Very' can operate on an adjective within a NP: (43) Munai sig über eng house very good the 'The very good house' It can also operate on a predicate (44) or a satellite (45): (44) wo sig ye-tanami igam-a he very me-surpass.SS stay-3s.DS 'He greatly surpasses me and (I)...' (45) Morogan eng sig erer qite qib igour bird the very on.top up.there go.SS they.stay 'Those birds fly very high' These adverbs show a striking similarity to what Quirk et al. (1985: 604) call "focusing subjuncts, which can draw attention to a part of a sentence as wide as the predication or as narrow as a single constituent of an element (such as a premodifying adjective in a noun phrase as subject or an auxiliary within a verb phrase)". They do more than draw attention, of course. They add, limit, or intensify. They appear to have a focusing function because they are the immediate scope of the negative, as (36) showed. The limited scope of this paper does not permit a systematic investigation of the adverbs listed in (14) and others. The restricted data do indicate that even the small class of adverbs is not homogeneous. Mom 'completely, already', qas 'only', and ar 'just, purposelessly' cannot operate on a level that would keep them outside the scope of negation. But of these three ar can also be used in adjectival position, e.g. munon ar eng 'the "idle" men' (i.e. those that are unmarked for a certain characteristic) and hence even as Head of a term: ar eng 'the one without the features'.

5. Proposition and illocution satellites If Papuan languages have such a restricted number of adverbs, in what way would proposition, or illocutionary, satellites be expressed? There is very little information in the descriptive grammars about these possibilities. It would seem

226

Ger P. Reesink

to me that adverbs functioning as attitudinal or illocutionary disjuncts are very rare, if not completely absent in Papuan languages. If illocutionary force is made explicit, it would normally be done in separate clauses, as in (46) Apand quraum,wo narau ba me is-au true I say he betelnut take not descend-Nom Ί speak truly, he has not taken the betelnut' Or, rather typical for the Usan culture, by some lengthy disclaimers: (47) Wa, ye misir boru ger qemerine isa gab wow, I thought bad onesay.ls.DS go.down.3s.DSsee.SS 'Wow, if (you) see that I say the wrong things...'

#

Perhaps the clause-initial bo 'again' in (30)d and gor 'also' in (31)a should be seen as functioning on the proposition level. Other candidates for this level may be inteijection-type adverbs, such as borumesis 'too bad, unfortunately', or the question word mai *what\ as in: (48) wurinou their

diab, mai, narau biga gab come up.SS what betelnut put.3s.DS see.SS

igo eng be.SS if 'If they themselves came up, for example, and were seeing that he put betelnut...' At this point, I would not have clear structural tests to assign either simple adverbs or adverbial expressions satellites status at the prepositional or illocutionary level.

6. Conclusion In the growing literature on the multitude of heterogeneous Papuan languages, there is very little information on adverbs and what are called particles. In the limited space of this article I have tried to bring together what little we do know. This suggests that the temporal and spatial satellites are clear and welldefined and that each language has a number of specific adverbs to express these predicate satellites (this level because they easily fall under the scope of negation and illocutionary operators such as interrogative and imperative). Most

Adverbs in Papuan languages

227

other "adverbial notions" are expressed through verbal complexes: verb + verb compounds or verb + nominal/adjective/adverb combinations. Then a rather small class of mostly monosyllabic, but certainly monomorphemic forms remain that are categorized as adverbs or particles. I have argued that they do not express satellites. But neither am I satisfied with their characterization as operators. The difference between grammatical and lexical means of specification is clear when we are dealing with inflectional affixation as opposed to free content words. It is less clear in the case of these minimal free words. This may be the reason why some linguists take their refuge to the term "particle" to name these forms. Their actual syntactic function does not seem to be different, however, from adverbs. And as adverbs they express operators (or satellites, if their lexical independence warrants such classification) on the predicate level as long as they are positioned between the first argument and the predicate. Their mutual differences suggest that a refinement in subclassification, such as e.g. Hoberg (1981) for German adverbs, is necessary. I would hope that this short account of adverbs in Papuan languages triggers more attention to a rather complex problem.

REFERENCES Anceaux, J.C. 1965. The Nimboran language. The Hague: Martinus Nijhoff. Bruce, L. 1984. The Alamblak language of Papua New Guinea. Pacific Linguistics C.81. Bunn, G. 1974. Golin Grammar. Workpapers in PNG languages, Vol.5. Ukarumpa: SIL. Davies, J. 1981. Kobon. Lingua Descriptive Studies, Vol.3. Amsterdam: North Holland. Dik, S.C. 1989. The Theory of Functional Grammar. Dordrecht: Foris. Dik, S.C. et al. 1990, The Hierarchical Structure of the Clause and the Typology of Adverbial Satellites. In: J. Nuyts, A.M.Bolkestein, C.Vet (eds.) Layers and levels of representation in language theory. Amsterdam: John Benjamins. Ernst,.T.B. 1984. Towards an integrated theory of adverb position in English. Bloomington, Indiana: I.U.L.C. Foley, W A 1986. The Papuan languages of New Guinea. Cambridge: University Press. Foremman V.M. 1974. Grammar of Yessan-Mayo. Ukarumpa: SIL. Giv6n, T. 1984. Syntax, Vol.1. Amsterdam: John Benjamins. Greenbaum, S. 1969. Studies in English Adverbial Usage. London: Longman.

228

Ger P. Reesink

Haiman, J. 1980. Hua, Α Papuan language of the Eastern Highlands of New Guinea. Amsterdam: John Benjamins. Hoberg, U. 1981. Die Wortstellung in der geschriebenen deutschen Gegenwartssprache. München: Hueber. McCawley, J.D. 1983. The syntax of some English adverbs. Chicago Linguistics Series 19, 263-282. McCawley, J.D. 1988. Adverbial NPs. Language 64, 583-590. Quirk, R., S. Greenbaum, G. Leech and J. Svartvik. 1985, A comprehensive grammar of the English language. London: Longman. Reesink, G.P. 1987. Structures and their functions in Usan. Amsterdam: John Benjamins. Roberts, J. 1987. Amele. London: Croom Helm. Schachter, P. 1985. Parts-of-speech systems. In: T.Shopen (ed.) Language typology and syntactic description. Cambridge: University Press, p. 3-61. Voorhoeve, C.L. 1965. The Flamingo Bay dialect of the Asmat language- The Hague: Martinus Nijhoff. Vries, LJ. de. 1989. Studies in Wambon and Kombai. PhD Diss. University of Amsterdam. Wilson, D. 1974. Suena grammar. Workpapers in PNG languages, Vol.8. Ukarumpa: SIL.

Functional prepositions

Henk van Riemsdijk Tilburg University

introduction2

0.

What is a syntactic phrase? It is the task of the theory of phrase structure, of Xbar theory2 for example, to answer such a question. But if we are honest, few answers have come forth and maybe the best thing we can say about the work on X-bar theory is that it has raised some interesting questions (without answering them). One of the main insights that X-bar theory incorporates and, to a certain extent, formalizes is that phrases are endocentric. But this is an insight due to structuralism and hence cannot be chalked up on the credit side. In addition, people have used the well-known but forever vague distinction between complements and adjuncts within X-bar theory and have endeavored to relate that distinction to a structural distinction in the hierarchical make-up of syntactic phrases. The evidence for such a structural correlate remains quite scant, however, even in languages with relatively fixed word order.

1

Thanks are due to Hans Bennis and Läszlö Maräcz for stimulating discussion.

2 See Chomsky (1970). For an overview of various issues as well as an extensive bibliography, see Muysken & Van Riemsdijk (1986).

230

Henk van Riemsdijk

A further notion worth mentioning in this context is that of specifier. This term is somewhat more authentic to generative grammar, perhaps, but (as a result?) it is also easily the most incoherent one. The term has variably been used to designate articles, possessor phrases, adverbs, the landing site for moved whphrases, subjects, and all kinds of (sub-)sets of these. In an attempt to sort out at least some of these problems, the notion of functional category has recently been introduced.3 Typical examples of elements analyzed in terms of this notion are articles (determiners), inflectional elements (including their free form spell-outs as auxiliaries) and subordinating conjunctions (complementizers). One property most analysts appear to agree on is that such elements are head-like, in other words that they are to be analyzed as Χ 0 / In view of the relatively substantial amount of work in this area, it is somewhat surprising that the identification of these functional categories in terms of categorial features has attracted virtually no attention at all. In previous work (Van Riemsdijk (1988)) I have speculated on this issue, and the present article pursues this speculation in the realm of prepositional phrases. The speculation is this. Verbs can take NPs, PPs, clauses, and maybe some other categories as their complements. Nouns are almost as free in their selectional behavior. Functional heads, however, entertain a much more intimate relationship with their complement or 'host projection'. D goes with NP, I (the inflectional node) with VP.5 Moreover, prototypical determiners very often seem to be morphologically nominal in nature in that one frequently observes a fair amount of similarity between the form of articles on the one hand and (pro-)nominal form classes such as pronouns, demonstratives, clitics and the like on the other. Similarly, inflectional elements, when spelled out as free forms, tend to be verbal in nature. While defective in many ways, for example, the English auxiliary is clearly a verbal element. We may call this hypothesis the categorial identity thesis, where by categorial identity we mean the categorial identity between the functional projection and the lexical host projection. This generalization should, I feel, be accorded a central position in any formulation of X-bar theory that incorporates the notion of functional head. One way of doing that is outlined in the next

3

See, among many others, Abney (1987), Fukui (1986) and Speas (1986).

4

Most researchers also conclude from this that these functional heads head their own projection. This is the version of the theory know as the DP-theory, meaning that the determiner (D) is a head projecting to a determiner phrase (DP) and taking the NP as its complement. See references cited above for details. In my own view this is not a necessary assumption and potentially an excessively powerful one. 5 I will not pursue the analysis of the complementizer projection (CP) here, precisely because CP does not have a lexical host projection. That is one of the reasons why I follow Emonds (1985) in assuming that the CP is essentially to be analyzed as a special kind of PP.

231

Functional Prepositions

paragraph. Then I will turn to the extension of the categorial identity thesis (CIT) to prepositional phrases.

1. A CIT version of X-bar theory Following Chomsky (1970) and Muysken (1983) I will assume two categorial features ([±N] and [±V]) and two features for distinguishing bar levels ([± PROJECTION] and [±MAXIMAL]).In addition, I will use the feature [±F] to distinguish between lexical and functional categories. We thus have the following set of features: (1)

{ {±N,±V} {±PROJ,±MAX} {±F} }

The categorial features define categories in the following manner. (2)

[+N.-V] [-N.+V] [+N.+V] [-N.-V]

= N, D = V, I = A = Ρ

The lexical and functional categories are distinguished by means of [±F], In order to better reflect the CIT, I replace the symbols D and I by the lower case letters corresponding to their respective host projections: η and v. (3)

[+N.-V.-F] [+N.-V.+F] [-N.+V.-F] [-N.+V.+F]

= = = =

Ν η V ν

(=D) (=1)

Bar levels are defined as in (3). (4)

[-PROJ,-MAX] X° = [+PROJ,-MAX] = X' [+PROJ,+MAX] = XP [-PROJ,+MAX] = reserved for non-projecting categories such as particles, clitics etc.

Projections can now be defined in terms of a small number of well-formedness conditions on the hierarchical dependencies between these features.

232

Henk van Riemsdijk

(5)

* -PROJ > +PROJ * -MAX > +MAX * -F > +F

(6)

Projection: vertical path through a tree such that that path satisfies the wellformedness conditions and such that all nodes on the path have the same categorial features

(">" = "dominates")

The canonical structures, disregarding left-to-right order, characterized by this set of definitions are as in (7). (7)

2.

Lp SPEC(?) L· - I- x° [r - Lr X° · · · ] ] ] ] ] (where ... stands for complements and adjuncts)

Uniform headedness

Some languages in the world are pretty, even from a theoretical point of view. For example, some languages like Japanese are uniformly head-final: nouns, verbs, adjectives, postpositions, they all follow their complements and adjuncts. And even the functional heads follow this scheme: determiners follow the noun, inflectional elements follow the verb and complementizers (if they are functional heads) follow the clause. But alas, many languages are not pretty. Dutch and German, for example, are verb-final but noun-initial. And as soon as we look at APs and PPs things get even messier: sometimes these categories are head-final, sometimes they are head-initial.6 Consider the following examples.7 (8)

gegenüber ihm vs. ihm gegenüber auf den Berg vs. den Berg hinauf

(opposite him) (up (onto) the mountain)

Probably we cannot maintain the idea that all languages fix the headedness parameter in a uniform way at the level of the grammar, even though Japanese apparently does.® But do we also have to give up the idea that headedness is at least uniform per category? Hopefully not. Certainly uniform headedness is a 6 In what follows, I will systematically disregard APs, both with respect to uniform headedness and with respect to the categorial identity thesis. This is left as a matter of future research. 7 I will limit myself to German examples whenever possible. The situation in Dutch (and in Swiss German, which I will occasionally invoke) is not essentially different. Ä It would be interesting to investigate the possibilities of analyzing Dutch and German as nounfinal underlyingly with some form of N-to-n movement accounting for the surface order, but at first sight such an enterprise appears to be rather hopeless.

Functional Prepositions

233

principle worth trying to maintain. What I wish to suggest is that an extension of the CIT to PPs makes it possible to maintain the uniform headedness principle for PPs in German and Dutch. Note that the proposed extension is not a stipulation designed to yield this result but a straightforward prediction following from the version of X-bar theory outlined above. Before going into the details of the analysis, however, we must investigate the properties of postpositional and circumpositional phrases in German (and Dutch) in more detail. This is the objective of the next section.

3. Postpositional and circumpositional phrases In addition to the usual prepositional phrases, German not only has postpositional phrases, as shown in (9) (and in (8) above), but also circumpositional phrases introduced by a preposition but closed off by another preposition-like element. Examples of the latter type are given in (10). (9)

den Berg hinauf (the mountain up(-P)) acc meiner Meinung nach (my opinion according-to) dat den Kindern zuliebe (the children for-the-benefit-of) dat den Verhältnissen entsprechend (the circumstances according-to) ihm gegenüber (him opposite / him towards) dat dem Verbot zuwider (the prohibition against) dat

dat

(10) auf den Berg herauf (on the mountain up(+Ρ)) acc hinter der Scheune hervor (behind the barn from(+P)) dat im Tal unten (in the valley down) dat gegen Ende hin (towards end (-P)) acc von Süden her (from south (+P)) dat auf mich zu (towards me to) acc unter der Brücke durch (under the bridge through) dat an seiner statt (in his place) gen um des Friedens willen (for the peace's sake) gen In many cases the postpositional element has the morpheme hin- or her- attached to it. These are deictic elements indicating movement away from or towards the point of reference (generally the speaker). In the glosses, these are are indicated as -P and +P respectively (where Ρ stands for proximate). In other Germanic dialects the postpositional element can be marked in different ways. In Swiss German, for example, the postpositional element is consistently marked by the affixation of a schwa.

234

Henk van Riemsdijk

(11)

Ziiritüütsch: uf NP uwe ab NP abe us NP use i NP ije dur NP dure

(up) (down, off) (out of) (into) (through)

acc dat dat acc acc

One important property of the postpositional element is that it is separable from the rest of the PP.9 Consider (12). (12) a.

b.

...dass er auf den Berg hinauf hat steigen wollen that he onto the mountain up has climb wanted ...dass er auf den Berg hat hinaufsteigen wollen

We should therefore ask ourselves whether the postpositional element could not be analyzed as a verbal particle which is never part of the PP. Several constituency tests argue against such an analysis. The postpositional element may sometimes be or become a verbal particle, perhaps through some process of incorporation, but genuine postpositional and circumpositional phrases do exist, as is shown by the fact that they occur in such canonical PP-positions as complement to Ν (cf. (13), complement to Ρ (cf. (14)), [PP with NP] (cf. (15)) and by the fact that they undergo wA-movement (or topicalization) (cf. (16)) and extraposition (cf. (17)). (13) a.

b.

der Blick [vom Fenster aus] the view from-the window out

c.

die Fahrt [den Fluss entlang] the trip the river along

(14) a.

9

der Weg [ins Tal hinunter] the way into-the valley down

bis [in die Niederungen hinunter] until in the lowlands down

b.

bis in die Niederungen

c.

bis hinunter

(compl. to NP)

(compl. to P)

See Van Riemsdijk (1978) for some discussion of the separability issue in Dutch.

235

Functional Prepositions (15) a.

[aus dem Gefängnis heraus] mit diesen Unschuldsengeln! out the jail out withthese angels-of-innocence

b.

[dort hinein] mit diesem Kram! there into with this junk

c.

[den Berg hinauf] mit euch! the mountain up with you (pi.)

(16) a.

b.

(17) a.

b.

[Meiner Meinung nach] hat er recht my opinion according-to is he right

(topicalization)

[Auf dem Turm oben] steht ein grosses Kreuz On the tower up stands a large cross Er hat nichts zu sagen gewusst [auf meine Frage hin] he has nothing to say known at my question (-P) (extraposition) Willst du dich nicht einmal umsehen [im Garten draussen]? don't you want to look around in-the garden outside

The problem of separability remains a vexing problem, of course, but one which we cannot pursue within the confines of the present article except to note the following descriptive generalization. (18) With subcategorized PPs, the postposition/particle generally must strand/separate/incorporate, i.e. adopt the behavior of a verbal particle, but with adjunct PPs pied piping is (sometimes) possible. This generalization is illustrated in the examples (19-22). (19) a.

der Fluss, an dem er entlang ging the river at which he along went

b. ?*der Fluss, an dem entlang er ging c.

der Fluss, an dem entlang ein Fusspfad gebaut werden wird the river at which along a footpath build be will

d. ??der Fluss, an dem ein Fusspfad entlang gebaut werden soll

236

Henk van Riemsdijk

(20) a.

die Frage, auf die hin der Hans starb the question upon which prt the John died

b. (21) a.

*die Frage, auf die der Hans hin starb der Standpunkt, von dem aus der Brief formuliert ist the point-of-view from which prt the letter worded is

b. (22) a. b.

*der Standpunkt, von dem der Brief aus formuliert ist die the die the

Kinder, denen zuliebe er etwas tut children, for the benefit of whom he does something (anything) Kinder, denen er etwas zuliebe tut children, for the benefit of whom he does something (specific)

This concludes the investigation of some basic syntactic properties of post-and circumpositional phrases. Let us now turn to their analysis.

4.

An analysis

In studying the question as to what is the proper analysis of post- and circumpositional phrases, we will concentrate on the choice among the following alternatives. (23)

First, (23a) can be eliminated as an option. Various considerations show that it is the leftmost P, not the rightmost prepositional element that governs the properties of the NP in between. This is true for case marking, as shown in (24) and for subcategorization, as shown in (25). (24) a. b. (25) a. b. c.

auf michocc zu unter der,.. Brücke durch auf dem Berg oben im Tal unten *NP oben/unten; *oben/unten NP

auf: acc zu: dat unter: dat/acc durch: acc on the mountain on-top in-the valley down

237

Functional Prepositions

(24) shows that it is the leftmost preposition that determines the case on the NP. Oben and unten in (25) are intransitive prepositions, as is seen in (25c). Nevertheless they occur in circumpositional PPs. Also, the deictic morpheme on the postpositional element is sometimes idiosyncratically selected by the preposition, as in 26). (26) a. zu alledem hinzu/'herzu (in addition to all of this) b. hinter der Scheune hervor/'hinvor (behind the barn from) Furthermore, assuming (23a) would imply that certain elements of the category Ρ have hybrid subcategorization or government properties, being sometimes leftand sometimes rightselecting: (27) a. unter der Brücke durch b. vom Fenster aus

durch [+P, _ N P , P P _ ] aus [+P, _ N P , PP ]

The position of modifying elements, which we generally expect at the beginning of a phrase, also argues against (23c): (28) a.

hoch (high) auf dem Berg oben •auf hoch dem Berg oben ??auf dem Berg hoch oben

b. 3m in das Zimmer hinein *in 3m das Zimmer hinein ??in das Zimmer 3m hinein

(3m into the room into)

Turning now to (23b), one major consideration arguing against this structure has to do with a general constraint against right branching phrases to the left of nouns. XPs to the left of nouns must be head final, as shown in (29). (29) a.

der seinem Besitzer treue Hund the his

b.

owner faithful dog

*der treue seinem Besitzer Hund

We find the same behavior with prenominal PPs. (30) a. b.

der reception Empfang nach Konzert the after dem the concert *der nach dem Konzert Empfang

238

Henk van Riemsdijk

But if true postpositions did exist, we would expect postpositional phrases to occur in prenominal position. However, they do not, as (31-33) demonstrate. (31) a.

die Reise den Berg hinauf the trip the mountain up

(32)

b.

*die den Berg hinauf Reise

a.

der Fusspfad am the footpath

(33)

See entlang

at.the lakealong

b.

*der am See entlang Fusspfad

a.

der Rückzug um des Friedens willen the withdrawal for the peace's sake *der um des Friedens willen Rückzug

b.

If we could say that all PPs are head-initial, regardless of their surface appearance, we would have an account for the general prohibition against prenominal PPs/" A second argument against (23b) comes from a locality condition on the doubling phenomena we find in pseudostranding constructions. Some German variants have stranding of prepositions with da. With locative prepositions, however, stranding is mostly impossible. Instead there is a pronominal copy of da (dr-) inside the PP. (34) illustrates this. (34) Da There

muss der Ball ('dr-)unter liegen must the ball under lie

The relation between the two pronominal elements is subject to locality restrictions. In particular, it cannot cross a PP-boundary. That is, a PP which is the complement to a preposition cannot contain a dr- which is connected to a da outside the containing PP.

The explanatory value of the account will depend, of course on what the head-to-head adjacency requirement on prenominal phrases is taken to follow from. The evidence from prenominal adjective phrases, participles and modal infinitives is interesting in this respect in that it suggests that adjacency is a condition on the morphological agreement relation between the head noun and the head of the prenominal phrase. The extension to prepositional phrases would then imply that such an agreement relationship exists for prepositional phrases as well, even though in the languages under discussion this agreement relation is not morphologically expressed.

239

Functional Prepositions (35)

*Da konnte man den Ton bis drunter hören there could one the tone until under hear

Observe now that the prepositional element in the post- and circumpositional phrases we are investigating is not subject to this locality requirement. (36) Da hätte er nicht drauf hinaufsteigen sollen There had he not thereup up climb should It seems reasonable, then, to eliminate (23b) as a possible analysis. As a result, we are left with (23c), the structure we set out to argue for. The proposal, in other words, is to consider the postpositional element in post- and circumpositional phrases as a functional Ρ (ρ), the prepositional counterpart of D (n) and I (v). The German paradigm shows that this analysis makes sense in yet another respect. One way of looking at functional heads is that they serve to express certain morphosyntactic features that are not normally expressed on the lexical head. German nouns, for example, do not carry a morphological mark for such features as gender or definiteness. These properties are expressed by the determiner. Similarly, auxiliaries serve to express various forms of modality when the main verb cannot do so itself. Along these lines we can say that the functional preposition serves to express certain locational dimensions where the lexical prepositional head does not do so itself. The following table constitutes a (partial) analysis of functional prepositions in terms of three main features (where the feature ORIENTATION has two subfeatures: IN and UP). DIR

OR

PROX

+ UP + / herauf/hinauf -UP herab,herunter/hinab,hinunter + /" oben + UP ο -UP 0 unten + + IN herein/hinein +/+ -IN heraus/hinaus +/+ IN (dr)in(nen) 0 -IN 0 (dr)aussen + her,zu/hin,weg 0 +/ο — +/Summing up, I propose that PPs in German and Dutch have the following structure. + +

-

-

-

(38) ^ [p ? P° NP ] p° ]

240

Henk van Riemsdijlc

In the case of postpositional phrases, I assume that there is movement from P° to p°. This movement, while lexically conditioned in many cases, is a regular instance of head-to-head movement of the type found in the verbal domain (V-to-I, V-toC) and, as argued recently, among others, in Longobardi (1990) in noun phrases (N-to-D movement). In the case of circumpositional phrases we can take the postpositional element to be base-generated. The dependency between the two prepositional elements is then to be regarded as an agreement relationship akin to that holding between the noun and the determiner with respect to such features as, say, gender and mass/count.

5.

A Hungarian analogue?

The analysis given above has the advantage of preserving the principle of uniform headedness. It does so, however at the expense of assuming that headedness may be different, even within one and the same projection, for the lexical and the functional head: Ρ is initial and ρ is final. This may well be the standard analysis of cases of mixed headedness. To conclude, let us look at a constuction which may well be profitably considered to be the mirror image case of German. There is little doubt that PPs in Hungarian are fundamentally postpositional. There are two main types of postpositions in Hungarian, called naked and dressed PPs by Maräcz (1989). In the same work, a very interesting fact is noted: apparently demonstrative NPs cannot occur inside dressed PPs in the same way that normal NPs do. Instead, the demonstrative element is combined with a copy of the dressed postposition. This is shown in the following examples.77 (39) a häz the house (40) a häz the house

az a häz this the house mögött behind

*az a häz mögött this the house behind

(41) a(z) mögött a häz mögött this behind the house behind (=behind this house) It is tempting to analyze the prepositional copy in Hungarian as a functional preposition on which the demonstrative feature is morphologically expressed in much the same way that the feature for ± proximity is expressed in German. We

The ζ of the demonstrative element is dropped in this construction due to a phonological process.

Functional Prepositions

241

could then say that Hungarian is P-final but p-initial, that is, Hungarian PPs would be the mirror image of German ones: (42)

[pp p° [ , , ρ Ν Ρ Ρ ] ]

Further research will have to determine whether this analysis is tenable. In conclusion, the above analysis of post- and circumpositional PPs in German (and Dutch) appears analytically correct, it offers a way to preserve the principle of uniform headedness and it supports the incorporation of the categorial identity hypothesis into X-bar theory.

REFERENCES Abney, S. 1987. The English Noun Phrase in its Sentential Aspect. Ph.D. dissertation, ΜΓΓ. Chomsky, N. 1970. Remarks on Nominalization. In: R. Jacobs & P. Rosenbaum (eds.). Readings in English Transformational Grammar. Waltham, MA: Ginn, p. 184-221. Emonds, J.E. 1985. A Unified Theory of Syntactic Categories. Dordrecht: Foris. Fukui, N. 1986. A Theory of Category Projection and its Application. Ph.D. dissertation, ΜΓΓ. Longobardi, G. 1990. Evidence for the Structure of Determiner Phrases and N-movement in the Syntax and in LF. GLOW Newsletter 24, 42-43. Maräcz, L. 1989. Asymmetries in Hungarian. Ph.D. dissertation, University of Groningen. Muysken, P.C. 1983. Parametrizing the Notion Head. The Journal of Linguistic Research 2, 57-76. Muysken, P.C. & H.C. van Riemsdijk 1986. Features and Projections. Dordrecht: Foris. Riemsdijk, H.C. van 1978. A Case Study in Syntactic Markedness. Dordrecht: Foris. Riemsdijk, H.C. van 1989. The Representation of Syntactic Categories. In: Proceeding? of the Conference on the Basque Language. Vol. I. Vitoria-Gasteiz: Basque Government Publications Office, p. 103-116. Speas, M. 1986, Adjunctions and Projections in Syntax. Ph.D. dissertation, ΜΓΓ.

On the internal structure of the Noun Phrase in Bahasa Indonesia

Sie Ing Djiang Institute for General Linguistics University of Amsterdam

0.

Introduction7

This paper is a partial study on the internal structure of the Noun Phrase in Bahasa Indonesia (Indonesian) within the framework of the Government and Binding theory as elaborated in Chomsky (1986), Fukui & Speas (1986) and other related work. We will show (in section 1) that in Indonesian the Determiner Phrase (DP) is head final. Then (in section 2) we will discuss modification structures involving elements which are traditionally called "Classifiers" and "Measure Words". We will argue that Classifiers and Measure Words are Functional rather than Lexical categories. Section 3 is concerned with the internal structure of nominal phrases containing other types of modifiers, including relative clauses introduced by the Complementizer yang 'that'. We will, however, not go into the question as to how such relative clauses are to be analyzed. In section 4 we will make a proposal as to how "Genitive Case" is assigned in Indonesian. Finally (in section 5), we will discuss the question as to whether "Quantifiers" constitute a separate syntactic category.

244

Sie Ing Djiang

1. The Structure of the Determiner Phrase Consider the following example (taken from Fokker 1951: 7): (1)

Percobaan itu gagal attempt that fail 'the attempt failed'

Within the theoretical framework adopted here the nominal phrase percobaan itu 'the attempt' in (1) has the following structure:

Ν

I

percobaan

itu

That is, the DP in Indonesian is head final. The S-structure of sentence (1) is as in (3):

N'

DET

INFL

V'

V percobaan

itu

t_

I

gagal

In (3) t is the trace of the D' which has been moved to the Specifier position of IP in order to receive (abstract) Case from INFL. The assumption is that Case percolates down to Ν (the head noun). Since in Indonesian the DP is head final, that is, the DET node is on the right hand side, we assume that the Specifier position in DP is also on the right hand side, as in (4), where SPEC is the Specifier position:2

On the internal structure of the Noun Phrase in Bahasa Indonesia (4)

245

DP

/

\ SPEC

D' / N'

/

N' \

Ν

\

DET

\DP) (DP)

As shown in (4), we assume that the structure of the DP in Indonesian is the mirror image of the structure of the DP in English (cf. Fukui & Speas 1986: 133). Note that in (2) and (3) the D-projection goes no higher than D'. We will have occasion to return to the Specifier position in DP in sections (4) and (5) below.

2.

Classifiers and Measure Words

In this section we will be concerned with the categorial status of what is called "Classifiers" and "Measure Words" in traditional grammars. In formal Indonesian a Numeral (Nu) cannot immediately precede a Noun (N). It must be followed by either a Classifier (CLAS) or a Measure Word (MW). That is, in a nominal phrase containing a Numeral the word-order is either Nu-CLAS-N or Nu-MW-N. In Indonesian Classifiers form a closed class. There are three general Classifiers and a small number (a score or so) of more specific ones. The three general Classifiers are orang, ekor, and buah. Orang is only used with Nouns with the feature [+human], ekor only with Nouns with the features [+animate, -human], and buah only with Nouns with the feature [-animate], as in the following examples: (5) a.

c.

seorang guru one-CLAS teacher 'a teacher'

b.

dua ekor ikan bawal3 two CLAS fish pomfret 'two pomfrets'

tiga buah rumah three CLAS house 'three houses'

An example of a more specific Classifier is butir, which is only used with Nouns with the features [-animate, + spherical, + small], as in (6):

246

Sie Ing Djiang

(6)

empat butir telur/kelereng/intan four CLAS egg marble diamond 'four eggs/marbles/diamonds'

Nominal phrases such as the following are ruled out: (7)

*seekor /'sebuah /'sebutir guru one-CLAS/one-CLAS /one-CLAS teacher 'a teacher'

The ungrammatical examples in (7) can be accounted for if we assume that a Classifier must agree in one or more features with the Noun with which it cooccurs, that is, if we assume that the Classifier orang has the feature (+human), the Classifier ekor the features [+animate, -human], the Classifier buah the feature [-animate], and the Classifier butir the features [-animate, + spherical, +small]. Measure Words include such words as kilo 'kilogram', dosirt 'dozen', keranjang 'basket', etc. Examples of nominal phrases containing Measure Words are: (8) a. duakilo /dosin /keranjang ikan bawal two kilogram /dozen /basket fish pomfret 'two kilograms/dozen/baskets of pomfrets' b. dua kilo /dosin /keranjang telur two kilogram /dozen /basket egg 'two kilograms/dozen/basket of eggs' The examples in (8) show that in the case of Measure Words there is no grammatical requirement that a Measure Word agree in some categorial feature(s) with the Noun with which it co-occurs: kilo 'kilogram', dosin 'dozen', and keranjang 'basket' can be used with a Noun with the features [+animate, -human] such as ikan bawal 'pomfret' (see 8a), as well as a Noun with the features [-animate, + spherical, + small] such as telur 'egg' (see 8b). Thus, it would seem that a Measure Word can freely co-occur with any Noun. This, however, is not true, as shown in the following examples: (9) a. dua keranjang ikan bawal /telur two basket fish pomfret /egg 'two baskets of pomfrets/eggs'

On the internal structure of the Noun Phrase in Bahasa Indonesia

247

b. dua keranjang 'guru /'rumah two basket teacher /house 'two baskets of teachers/houses' The examples in (9) show that, in order to be able to co-occur with a specific Noun, a Measure Word must be "appropriate" in some sense. We will leave open the question as to how this semantic restriction on the use of Measure Words is to be encoded in the grammar. What we are concerned with here is the question: what is the internal structure of a DP containing a Classifier or a Measure Word? Note first that both Classifiers and Measure Words are not Common Nouns. To see what we mean, take, for instance, the Classifier elcor [+animate, -human] and the Measure Word keranjang 'basket'. The Classifier ekor is homophonous with the Common Noun ekor 'tail', which takes the Classifier batang [-animate, + cylindrical], and the Measure Word keranjang is homophonous with the Common Noun keranjang ^basket', which takes the Classifier buah [-animate]. Thus, we have the following paradigm: (10) a.

b.

(11) a.

b.

dua batang ekor lembu two CLAS tail ox 'two ox-tails (used for soup)' dua ekor lembu two CLAS ox 'two oxen' dua buah keranjang sampah two CLAS basket rubbish 'two waste-baskets' dua keranjang sampah two MW rubbish 'two baskets of rubbish'

Ekor lembu 'ox-tail' in (10a) and keranjang sampah Vaste-basket' in (11a) are Common Nouns, but ekor in (10b) is a Classifier and keranjang in (lib) a Measure Word. Suppose now that, in addition to the Functional categories COMP, INFL, and DET, Indonesian has two other Functional categories, namely, CLAS (Classifier) and MW (Measure Word). Like DET, CLAS and MW select a unique complement: DET selects N' as its complement, whereas CLAS and MW select Nu' (Numeral) as their complement. Under this assumption, the

248

Sie Ing Djiang

nominal phrases sebutir telur 'an egg' and sekeranjang telur 'a basket of eggs' have the internal structures (12a) and (12b), respectively: D'

(12) a. N' CL'

/

DET Ν

e

CLAS

Iju' Nu se-

butir

b.

telur

D' N' MW' / Nu'

I

DET Ν

e_

\ MW

Nu se-

keranjang telur

The structures (12a) and (12b) show that a Noun may be modified by a Functional category projection headed by either the Functional category CLAS or the Functional category MW. Here modification is structurally defined as in Speas (1986: 85): (13) X modifies Υ -» Χ Λ Y form a constituent dominated by a projection of Y (order irrelevant) The difference between a CLAS and a MW is that a CLAS must agree in one or more features with the head noun, whereas a MW need not. In (12a) and (12b) the Case assigned to the D' percolates down to the head noun (and also to the Numeral, if we assume that Numerals constitute a subclass of Nouns; see section 6). The structures (12a) and (12b) automatically accounts for the fact

On the internal structure of the Noun Phrase in Bahasa Indonesia

249

that a Noun cannot take more than one CLAS or MW: phrases such as *sebuah butir telur 'one CLAS CLAS egg' or *sekilo keranjang telur 'a kilogram basket of eggs' are ill-formed. An issue which we will not pursue relates to the assumption that a Determiner Phrase may be headed by an empty (i.e., phonetically null) DET ~ an assumption which has direct bearing on the problem of "Theta-binding" (in the sense of Higginbotham 1985) in nominals (cf. Fukui 1986:178, footnote 20). Given this assumption, the DET-position in (12a) and (12b) is filled by a phonetically null determiner (e), which presumably may have features such as (+/-definite), etc. Now consider the following two examples (taken from Fokker 1951: 24 and 141, respectively): (14) Seorang telah menikamkan goloknya ke leher babi one-person Perfective stick machete-he to neck pig 'Someone had already stuck his machete into the neck of the pig' (15) Juru-juru rawat membawa seorang sakit masuk nurse-nurse take a person sick enter "The nurses carried a (sick person=) patient inside' Apparently the word order in the nominal phrases seorang 'a person = someone' in (14) and seorang sakit 'a patient' in (15) is Nu-N, that is, these nominal phrases apparently are exceptions to the rule that a Numeral cannot immediately precede a Noun. This apparent irregularity can be accounted for if we assume (i) that orang 'person' and orang sakit 'sick person = patient' regularly take the Classifier orang [+human], and (ii) that a sequence of identical phonological forms such as orang orang is forbidden by a "rule of cacophony" in PF, which has the effect of deleting one of the two instances of orang (for the term "rule of cacophony" see Gross 1967). Given these assumptions, the nominal phrases seorang and seorang sakit have the underlying structure (16):

250

Sie Ing Djiang

(16) N'

κ

D' DET

κ Ν

CL'

/

Nu'

e

CLAS

Nu

I

Sl se

orang orang orang orang

sakit

The rule of cacophony deletes the first instance of orang. An argument in favor of the structure (16) is the fact that in older written Malay there is no such rule of cacophony, as attested in the following example taken from Hopper (1986: 314), where the Classifier orang is not deleted: (17) Maka ku-dapati ada di-tengah rumah ada sa-orang orang and I-find be at-center house be one-CLAS person Pelekat yang bernama Abdul Satar tengah makan Pelekat COMP be called Abdul Satar Durative eat 'And I discovered that there was a man from Pelekat by the name of Abdul Satar in the middle of the room engaged in eating' Finally, consider the following example: (18) Aku hanya membeli ayam seekor dan beras dua liter I only buy chicken one-CLAS and rice two liter Ί only bought one chicken and two liters of rice' The nominal phrases ayam seekor and beras dua liter have the structures (19a) and (19b), respectively:

On the internal structure of the Noun Phrase in Bahasa Indonesia (19) a.

/ Λ DET I CL' e

D'

b. .Ν'

Ν' Ν

ayam

Nu'

Nu

Nu beras dua

ekor

DET

\ MW'

Ν

/ \ Nu' CLAS

se

251

MW

liter

The above structures show that in the case of modification the modifier may follow the modificand. That is, we assume that (at least in the case of modification by either a Classifier or Measure Word phrase) the modifier may either precede or follow the modificand (compare (19a-b) with (12a-b)). This, however, would imply that Nominal phrases in Indonesian may be either headinitial (as in (19)) or head-final (as in (12)). Alternatively, we could assume that Nominal phrases in Indonesian are underlyingly head-initial, as in (19a-b), and that there is a rule (an instance of head-to-head movement) which moves Ν to DET (cf Longobardi 1990, where he argues for N-to-D movement in Roman syntax, and van Riemsdijk 1990, where in his analysis of the German PP he assumes a head-to-head movement rule which moves Ρ to the head position of a Functional (Prepositional) category). Under this analysis the strings sebutir telur 'an egg' in (12a) and sekeranjang telur 'a basket of eggs' in (12b) are not base-generated but are the result of N-to-DET movement. Whether such an analysis is tenable, has to be determined by further research.

3.

Other types of modifiers

In addition to Classifier and Measure Word phrases, nominal phrases may contain other types of modifiers, as exemplified by the relevant nominal phrases in the following examples: (20) Tidak not

seorang-pun one-person-even

dapat mengubah can alter

yang tersurat pada COMP predestined in

azalnya (Fokker 1951: 55) life-span-he 'Nobody can alter his predestined life-span' (literally: Not even one person can alter what is predestined in his life-span)''

252

Sie Ing Djiang

The nominal phrase tidak seorang-pun 'not even one person = nobody' in (20), with the underlying structure (21), shows that an adverb and the negative element tidak may function as modifiers: D'

(21) N' N'

/ \

CL'

/

Nu' Χ Neg tidak

/ \

κ

\ DET

Adv

Ν

\

CLAS

\ Nu se

orang

orang pun

The first instance of orang in (21) is deleted in PF (cf (16)). (22) Kapalyang besar itu kapal pengangkut (Slametmuljana 1969:274) ship COMP big that ship freighter 'That big ship is a freighter' The nominal phrase kapal yang besar itu 'that ship which is big = that big ship' in (22) contains a clausal modifier in the form of a restrictive relative clause introduced by the Complementizer yang 'that' and has the structure (23):

Pending further study we assume that a relative clause introduced by the Complementizer yang is C rather than CP (C'). Note that relative clauses can be "stacked", as in (24a), where the relevant nominal phrase has the structure (24b):

On the internal structure of the Noun Phrase in Bahasa Indonesia (24) a.

Banyaklah negeriyang bam yang bertemu many-Emphatic land COMP new COMP discovered

253 oleh by

orang Portugis (Fokker 1951: 216) person Portuguese 'Many were the new lands which were discovered by the Portuguese'5 b.

C negeri

yang bertemu oleh orang Portugis

yang barn

(25) Orang di dusun itu sangat miskin (Slametmuljana 1969: 237) person in village that very poor 'The people in that village are very poor' The structure of the nominal phrase orang di dusun itu in (25), which contains a prepositional modifier, is as in (26): (26)

D'

/

\ DET

N' / Ν

\

I Ρ'

e

/

N' I Ν orang

di dusun

D'

\ DET itu

An example of a nominal phrase containing a so-called "Bare-NP adverb" is keadaan sekarang 'situation this moment' in (27), which has the underlying structure (28):

254

Sie Ing Djiang

(27) Keadaan situation

sekarang this-moment

memaksa mesti force must

berpikir think

dalam-dalam (Fokker 1951: 136) deep-deep 'The situation at this moment forces one to think deeply* D'

(28) N'v

N/

\ N'

DET I

/V

DET

Ν keadaan sekarang In (27) sekarang is analyzed as a "Bare-NP adverb" (i.e. a Noun) meaning 'this moment', not as an adverb meaning 'nowadays'. This analysis is supported by the fact that, as a "Bare-NP adverb", sekarang can be further modified, as in: (29) Aku mengharap I expect sekarang this moment Ί expect that them to reply

supaya jawabitu COMP reply that

diberikan be-given

ini juga (Fokker 1951: 172) this too the reply will be given at this very moment = I expect at this very moment'

Sekarang ini juga 'this very moment' in (29) has the structure (30): DP

(30) D'

Adv

/

\ DET

sekarang

ini

N' I Ν

juga

On the internal structure of the Noun Phrase in Bahasa Indonesia

255

Note that (30) is a full DP (D") where the Specifier position is filled by a modifying Adverb. On the question as to how "Bare-NP adverbs" are assigned Case, see Larson (1985). Finally, consider the following example (taken from Fokker 1951: 185): (31) sehelai selimut kecil compang-camping one-CLAS blanket small tattered 'a small tattered blanket' The structure of the nominal phrase (31) is as in (32): D'

(32) κ N'

/

N'

N'

κ \

/

CL'

κ \

Nu'

\

DET

\A

A

\Ν

CLAS

Nu se

helai selimut kecil compang-camping

We assume, with Speas (1986: 67-68), that adjectival modifiers are A rather than A'. Summarizing so far, we have seen that in Indonesian a nominal phrase may contain various types of modifiers, including a Classifier phrase, a Measure Word phrase, Adverbials (including the Negative element), relative clauses, a prepositional phrase, a "Bare-NP adverb", and Adjectives.

4. The Genitive Case Consider the following example (taken from Fokker 1951: 31): (33) A tap rumah itu seng roof house that corrugated-iron 'The roof of that house is (made of) corrugated iron'

256

Sie Ing Djiang

What is the internal structure of the "Genitive" construction atap rumah itu 'the roof of that house' in (33)? A possibility is to assume, with Lumsden (1989), that, like Haitian Creole, Indonesian has a phonetically null DET (0) which assigns Genitive Case to the Specifier position of DP. Under this assumption, the nominal phrase atap rumah itu 'the roof of that house* has the S-structure (34):

N'

/

/ \ I i atap Ν

\ D' t_

DET

/

N'

I IΝ

0

κ

rumah

\DET itu

In (34) t is the trace of the D' which has been moved to the Specifier position of DP in order to receive Genitive Case. The assumptions here are (i) that a Noun like atap 'roof in (34) cannot assign Case to its sister D', and (ii) that the empty DET (0) assigns Genitive Case rightward. The latter assumption is contrary to the assumption in Fukui & Speas (1986: 140) that all Functional categories assign Kase (which includes Nominative and Genitive Case) leftward, and that this is one of the properties which distinguish Functional categories from Lexical ones (which assign Case rightward). Another possibility is to assume (i) that in Indonesian a Noun can assign structural Genitive Case to its right sister D', and (ii) that a Functional category cannot assign Kase rightward; i.e., that, unlike Haitian Creole, Indonesian has no empty DET which assigns Genitive Case to the Specifier position of DP. Under this assumption, the nominal phrase atap rumah itu 'the roof of that house' has the structure (35):

On the internal structure of the Noun Phrase in Bahasa Indonesia

N'

257

DET

I

Ν atap

rumah

itu

In (35) the Noun atap 'roof assigns structural Genitive Case to the D' rumah itu 'that house', and there is no movement to the Specifier position of DP. Now consider the following example: (36) Pembakaran toko-toko oleh kaum perusuh itu burning store-store by group rioter that menimbulkan kerugian besar cause loss great 'The burning of stores by the rioters caused heavy losses' The relevant nominal phrase in (36) has the underlying structure (37): DP (37)

/

/

D'

N' N' Ν

D'

\DET

SPEC

itu

D'

pembakaran toko-toko kaum perusuh In (37) the highest DET is filled by the demonstrative itu 'that, the', which does not assign Kase to the Specifier position of DP. This being the case, neither the D' toko-toko 'stores' nor the D' kaum perusuh 'the rioters' can move to the Specifier position to receive Genitive Case. The D' kaum perusuh gets its Case

258

Sie Ing Djiang

by a rule of oleh "by'-insertion. The question here is: how does the lowest D' (toko-toko 'stores') get its Genitive Case? In his analysis of Haitian Creole, Lumsden assumes that a Proper Noun or a Noun expressing a familial relation gets its Genitive Case in an exceptional way, without having to move to the Specifier position of DP, as in (38a), which has the structure (38b) (cf Lumsden 1989: 86): (38) a.

Jou£t Jan yo toy John pi. John's toys

b.

DP

NP

D"

I

yo

N'

Jou&t

I D* NP

I

N'

D"

0

I N° Jan The structure (38b) is parallel to the structure (37) in that the lowest DP (/an) cannot move to the Specifier position of the highest DP in order to receive Genitive Case. Now, even if we assume with Lumsden that in Indonesian, too, a Proper Noun or a Noun expressing a familial relation gets its Genitive Case in situ, example (36) ought to be ungrammatical, since in (37) the D' toko-toko 'stores' is neither a Proper Noun nor a Noun expressing a familial relation, and therefore cannot receive Genitive Case in situ. In order to account for the grammaticality of (36), we assume that in Indonesian a Noun may assign structural Genitive Case to its right sister D' (which is consistent with the general assumption that Lexical categories assign Case rightward) and that, unlike Haitian Creole, Indonesian has no empty DET which assigns Genitive

On the interned structure of the Noun Phrase in Bahasa Indonesia

259

Case rightward to the Specifier position of DP (i.e., there is no need to assume that Functional categories can assign Kase rightward). Given these assumptions, the nominal phrase atap rumah itu 'the roof of that house' in (33) has the Sstructure (35) rather than the S-structure (34). In (37) the D' toko-toko 'stores' is assigned Genitive Case by the Noun pembakaran 'burning'. Our analysis of the Genitive construction automatically accounts for the fact that the Indonesian "passive" nominal phrase pembakaran toko-toko oleh kaum perusuh itu 'the burning of stores by the rioters' in (36) has no "active" counterpart. That is, in Indonesian there is no structure equivalent to English the rioters' burning of stores, for the reason that under our analysis the D' kaum perusuh 'the rioters' in (37) cannot receive Genitive Case. Note that Genitive Case can be assigned to nominal phrases with different theta-roles. In (35) (atap rumah itu 'the roof of that house') the D' rumah itu 'that house' has a theta-role something like 'Possessor', but in (37) the D' toko-toko 'stores' has the theta-role 'Theme'. Under our assumptions, the relevant nominal phrase in (39a) (taken from Fokker 1951: 24) has the structure (39b): (39) a. Anaknya yang perempuan itu belajar pada sekolah rendah child-he COMP female that study at school low 'His daughter is in elementary school' (literally: that child of his who is female studies at the elementary school)

anak

N'

DET

Ν

e_

nya

In (39b) the pronoun nya 'his' is assigned Genitive Case by the Noun anak 'child'. The structure (39b) further shows that restrictive relative clauses are generated within the highest D' (see also (24b)). Now consider the following example (taken from Dardjowidjojo 1978: 340):

260

Sie Ing Djiang

(40) Organisasi organization

ini, yang dinamakan Budi Utomo dan this COMP be-named Budi Utomo and

didirikan pada tahun 1908, akhirnya mengalami suatu be-established at year 1908 finally undergo one perubahan orientasi change orientation 'This organization, which was named Budi Utomo and was establised in 1908, in the end changed its orientation' The relevant nominal phrase in (40) has the following structure: DP

(41) κ D' Ζ" N' I Ν

C DET I ini

yang dinamakan Budi Utomo dan didirikan pada tahun 1908

orgamsasi The structure (41) demonstrates that a non-restrictive relative clause is generated in the Specifier position of DP (D").

5.

Quantifiers

The general assumption is that elements such as se- 'one', dua 'two', etc. (see section 3 above) are considered to belong to the class of Quantifiers, that is, elements which are interpreted as Quantifiers at LF. The question here is: do Quantifiers constitute a syntactic category? In section 3 we have shown that, at least in Indonesian, such elements as se- 'one', dua 'two', etc., constitute a syntactic category, namely, that of Numerals, which can syntactically be defined as the class of elements which can serve as the unique complement of either the Functional category CLAS or the Functional category MW. Let us assume that Numerals form a subclass of Nouns. Specifically, let us assume that a Numeral has the features [+N, -V, +Numeral, + Quantifier], where the feature [+Numeral] indicates that an element with this feature has the syntactic property of being able to serve as the

On the internal structure of the Noun Phrase in Bahasa Indonesia

261

unique complement of CLAS or MW, and the feature [+Quantifier] indicates that an element with this feature is interpreted as a Quantifier at LF. Now consider the following two examples (taken from Fokker 1951: 46 and 144, respectively): (42) Kedua anak itu mandilah dengan riang both child that bathe-Emph cheerfully 'Cheerfully, both the children had a bath' (43) Segala penduduk telah lari ketakutan all inhabitant Perfective flee in fear 'All (of) the inhabitants had fled in fear' The questions here are (i) what is the internal structure of the nominal phrases kedua anak itu "both the children' in (42) and segala penduduk 'all of the inhabitants' in (43), and (ii) what is the categorial status of Quantifiers such as kedua 'both' and segala 'all'? Let us start with the second question. Nominal phrases such as (44a) and (44b) are ill-formed: (44) a.

b.

"kedua orang anak itu both CLAS child that 'segala orang all CLAS

penduduk inhabitant

That is, (44a) cannot mean 'both the children', nor can (44b) mean 'all of the inhabitants'. This indicates that Quantifiers like kedua *both' and segala 'all' are not Numerals. Let us assume (i) that Quantifiers such as the above form a subclass of Nouns (let us call it Nq) with the features [+N, -V, -Numeral, + Quantifier], and (ii) that Nq can assign structural Genitive Case. Under these assumptions the structures of kedua anak itu 'both the children' in (42) and segala penduduk 'all of the inhabitants' in (43) are as in (45a) and (45b), respectively:

262

Sie Ing Djiang

(45)

a.

D'

κ \ / \ I Nq'

Nq

D'

b.

D'

e.

\

Nq'

DET

DET Nq ,

N'

DET

N'

Ν

itu

Ν

i DET

kedua anak

segala penduduk

In (45a) and (45b) Nq assigns Genitive Case to its sister D\ Our assumption that Nu and Nq form distinct subclasses of Nouns with the properties and features as discussed above not only obviates the necessity of assuming a separate syntactic category of Quantifiers, but also accounts for the different syntactic behavior of Nu and Nq.

6.

Conclusion

We have shown that, given the theoretical framework adopted in this paper, in conjunction with the additional assumptions (i) that the DP in Indonesian is head final, (ii) that Classifiers and Measure Words are Functional categories, and (iii) that in Indonesian Nouns can assign structural Genitive Case, a variety of nominal structures can be accounted for in a principled way. We have further shown that, in Indonesian at least, there are two classes of Quantifiers which syntactically belong to two subclasses of nouns (namely, Nu and Nq) with distinct properties.

NOTES

1. The author is grateful to Henk van Riemsdijk and Pieter Muysken for helpful comments on an earlier version of this paper. Only the author is responsible for all remaining errors. 2. For a similar assumption, see Lumsden's (1989) analysis of Determiner phrases in Haitian Creole. 3. We assume that there is a productive morphological rule in Indonesian for the formation of nominal compounds such as ikon bawal 'pomfret' (literally 'pomfret fish'), ikan tongkol 'tuna fish', etc. 4. The phrase yang tersurat pada azalnya in (20) is a so-called "headless relative" meaning "what is predestined in his life-span'.

On the internal structure of the Noun Phrase in Bahasa Indonesia

263

5. In (24a) "Subject-Predicate Inversion" (discussed in Sie 1989) has applied, and the predicate banyak 'man/ contains the Emphatic particle lah.

REFERENCES Chomsky, N. 1986, Barriers. Cambridge, Massachusetts: the MIT Press. Dardjowidjojo, S. 1978, Sentence Patterns ofIndonesian. Honolulu: the University Press of Hawaii. Fokker, AA. 1951, Inleiding tot de Studie van de Indonesische Syntwäs (Introduction to the Study of Indonesian Syntax). Groningen-Jakarta: J.B. Wolters. Fukui, Ν. 1986, A Theory of Category Projection and Its Applications. Ph.D. dissertation, ΜΓΓ. Fukui, Ν. & M. SPEAS 1986, Specifiers and Projection. ΜΓΤ Working Papers in Linguistics 8,128172. Gross, M. 1967, Sur une rögle de cacophonie. Langages 7:105-119. Higginbotham, J. 1985, On Semantics. Linguistic Inquiry 16, 547-593. Hopper, PJ. 1986, Some Discourse Functions of Classifiers in Malay. In: C. Craig (ed.), Noun Classes and Categorization. Amsterdam/Philadelphia: John Benjamins, p. 309-325. Larson, R.K. 1985, Bare-NP Adverbs. Linguistic Inquiry 16:595-621. Longobardi, G. 1990, Evidence for the Structure of Determiner Phrases and N-movement in the Syntax and in LF. GLOW Newsletter 24, 42-43. Lumsden, J.S. 1989, On the Distribution of Determiners in Haitian Creole. Revue qutbicoise de linguistique 18.2, 65-94. Riemsdijk, H.C. van 1990, Functional Prepositions. In this volume. Sie, I.D. 1989, The Syntactic Passive in Bahasa Indonesia: a Study in Government-Binding Theory. Doctoral dissertation, University of Amsterdam. Slametmuljana, 1969, Kaidah Bahasa Indonesia (The Rules of Bahasa Indonesia). Ende (Flores): Nusa Indah. Speas, M. 1986, Adjunctions and Projections in Syntax. Ph.D. dissertation, MIT.

Deverbal nominalization in Sranan: a search for regularities

Norval S. Smith Institute for General Linguistics University of Amsterdam

0.

Introduction

The purpose of this brief article is to cany out a preliminary exploration of one type of reduplicative process in Sranan, the main Creole language of Surinam. This nonproductive process derives nouns from reduplicated verb roots. The qualification "preliminary" is essential, since the data on which this study is based are all from secondary sources. We will try to determine whether this process involves any morphosyntactic or phonological regularities. Our motivation for doing this is not so much synchronic as diachronic, in particular in connection with the long-standing debate around the question of the role of substrate languages in creole genesis. In the case of the creole languages of Surinam it has been claimed by Smith (1987) that there are three candidates for the role of significant substrate language. These are, in order of likely importance: Gbe (more familiar under the term Ewe/Fon), spoken in the coastal areas of Ghana, Togo, Benin, and Eastern Nigeria; Kikongo, spoken in the western parts of the two Congos, and of Angola; and thirdly Twi,

266

Norval S. Smith

spoken in Southern Ghana. We will not examine these languages in the course of this article but intend to proceed to such an examination in future work. Note that we assume here that the equivalents of English adjectives are verbs in the Surinam Creoles, following the majority of recent work. Compare in this respect the various articles in Alleyne (ed.) 1987. The total number of deverbal reduplications resulting in nouns is fairly limited, totalling under thirty forms.

1. The data We will start out by listing those forms that potentially seem to belong here together with their meanings in (1). Table (1) Verb Meaning Ν (20 c.)

Ν (1856)

Ν (1778)

Meaning

arari rake ba'bari bari bari tumult babari bronbron bronbron burnt food-crust dorodoro dorodoro sieve djompodjompo djompo djompo grasshopper fomföm fumfumm blow, beating fonfon freyfrey fly grater gritgriti gritigriti kamkam kankan kamkamm comb kap'kapoe machete kofoe kofoe cuff koskosi koskosi kossikossi curses kosokoso kossökossö cough (koso absent) -14 koti cut kottikotti slice scrapings -15a krabu scratch,scrape krabkrabu krabbo-krabbo rake -15b krabu scratch,scrape krassikrassi rash -16 krasi scratch kraskrasi kraskrasi krasikrasi -17 lau crazy laulau folly moi beautiful moimoi mo'möoi -18 finery mooi-möoi nainai nanai nanai needle -19 nai sew

-1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13

ari draw bari shout bron burn doro sieve djompo jump fon beat frey fly griti grate kan comb kapu chop kofu cuff kosi curse koso cough

-

267

Deverbal nominalizations in Sranan: a search for regularities Table (1) continued Verb Meaning Ν (20 c.) njanjan

-20

njan eat

-21 -22 -23

saka let down sakasaka seki shake sekseki sibi sweep sisibi

-24 -25a -25b -26

taki talk tai tie tai tie tja(ri) carry

taktaki ta(i)tai titei tjatjari

-27

wai

wawai

fan

Ν (1856)

Ν (1778)

njamnjam

jamjam food janjam sakkasakka dregs

saka-saka rattle sisibi sisibi sibisibi taktaki takkitakki tetöi tja'tjari tjari tjari wawai waiwai

Meaning

broom

gossip, nonsense bundle tetfei, tetöh rope tjarritjarri headpad

waiwai

fan

(N = noun) The historical sources we have employed are for the 19th century, Wullschlägel 1856, and for the 18th century, Schumann 1778. 1.1.

Problem cases

The first question we have to ask ourself is whether all these cases are what they are made out to be - reduplicative nominal derivatives based on a verbal root. In at least two cases there is reason for believing that this is not the case, and in a third case there is some doubt as to the correct status. The relationship between ari and arari is certainly not that implied above, an means "draw, pull", and arari "a rake", arari is quite certainly derived, not from ari, but from the iterative verbal reduplicative formation arari "to rake", with the implication of repeated drawing of the rake towards one. The noun arari is then derived by 0-derivation from the verb arari. We can assume that taktaki is a similar case, taki means "speak, say, talk" while taktaki means "gossip, chatter, nonsense". However, we also have a verb taktaki "to chatter, etc." which would seem a much more probable source - via 0-derivation once more - for the noun taktaki. Once again this would involve an iterative formation with the implication of repeated talking. The doubtful case involves doro "to sieve". In addition to the noun dorodoro, we also have another iterative verbal formation dorodoro, also given with the

268

Norval S. Smith

dictionary meaning of "to sieve". Clearly sieving is a process that easily lends itself to repetition, so that it is at least a possibility that the noun dorodoro might be derived from the reduplicated form of the verb. We will assume, however, that this is not so, since the noun itself does not seem necessarily to possess any kind of repetitive sense. 1.2.

Argument structure

Let us now turn to a consideration of the argument structure of the verbs involved. In a number of cases we might want to assign different argument structures to some of the verbs concerned, because of the extensive multifunctionality - also within grammatical categories - to be met with in Sranan. In such cases we have felt justified in selecting the argument structure that seemed to us to be closest to the sense of the relevant reduplicated nominal. At this juncture it requires to be stated that there is a considerable variety of ideas concerning the argument structure to be assigned to various verbs. For instance take the verb eat in English. This occurs either with one argument or two: (2)

a. b.

John ate the potatoes, John ate.

It is possible to approach this situation in two ways. Either we say that there are two verbs eat in English with the following argument structures (corresponding to (2)a. and b. above): (3)

a. b.

(Theme)Agent ( )Agent

or we assume that there is only one verb eat with the argument structure of (3)a. The theme would then be understood, or saturated. Equally unclear is the situation regarding certain possible prepositional arguments, such as instrumental expressions, as for example: (4)

a. b.

I cut the bread with a knife, I cut the bread.

Is it correct to speak of a three-argument verb in this case, or not? Is the instrumental expression with a knife merely an adjunct, or is the verb cut strictly subcategorized in respect of Agent, H e m e , and Instrument? It would seem reasonable to speak of strict subcategorization in this case. The instrument, if

Deverbal nominalizations in Sranan: a search for regularities

269

expressed, requires to be inanimate, as well as possessed of certain semantic features, such as [+THIN OBJECT], these being dependent on the particular verb concerned. So we have the following judgements: (5)

a. b. c. d. e. f. g· h.

j j J

?* « • *·

I I I I I I I I

sliced sliced sliced sliced sliced sliced sliced sliced

the the the the the the the the

bread bread bread bread bread bread bread bread

with with with with with with with with

a knife. a wire. a chain-saw. my hand. my foot. a ball. a ship. the truth.

Added to this we have the same possibility as with eat of not actually expressing an instrument with such verbs. The instrument is however always understood in such cases. Note that the situation regarding the instrument in the above case is completely different from that regarding other PP's occurring with verbs. Consider the following cases: (6)

a. I went to school. b. I went to school with Bill.

Here we have a PP in the b. sentence expressing accompaniment. This is absent in the a. sentence, and is most certainly not understood in such a case. There is just no accompaniment involved in the first sentence, while there is in the second. In the accompaniment phrase we clearly do not have anything that we could refer to as an argument of go. Let us turn now to a consideration of the argument structure of the verbs in (1). First of all we must mention the problem of the θ-roles of monovalent verbs. Assuming that we do not want to assume that the argument structure of verbs with specific meanings is universally determined, then we require independent grammatical evidence to assist in determining what the θ-role of a particular verb might be, in addition of course, to an awareness of the semantics involved. This problem is at its most serious with respect to verbs with a sole argument. On what grounds can we determine whether this argument is functioning as a Theme or an Agent? For instance, isflya verb that takes a theme or an agent. Considering that the context is 18th/19th century Surinam, and restricting ourself to the world of reality,

270

Norval S. Smith

the answer would seem to be that in purely semantic terms, we have a Theme here. Grammatical evidence bearing on this issue is absent as far as we are aware, although this is not to say that it does not exist. Other cases where even the semantics do not provide any clearcut decision would be jump and cough, "jumping" and "coughing" may be deliberate actions, they may also be inadvertent reactions to stimuli of various kinds. For reasons such as these, we will avoid making any decisions as to the particular θ-role of the single argument of monovalent verbs, and lump all these cases together as ( )Arg. Verb

Meaning

koso lau moi bari djompo frey

cough crazy beautiful shout jump fly

Turning to bivalent verbs, we have cases with the structure (Th)Ag, i.e. verbs with a Theme as internal argument, and an Agent as external argument. Note that we have considered bodily actions to be non-instrumental, e.g. "beat", "cuff', "scratch", etc.: Meaning Verb bron fon kofu kosi krabu krasi njan seki saka tjari

burn (causative) beat cuff curse scratch scratch eat shake let down (causative) carry

Finally we have a group of trivalent verbs with the argument structure (Th,Inst)Ag:

Deverbal nominalizations in Srancuv a search for regularities Verb

Meaning

doro griti kan kapu koti nai sibi tai wai

sieve grate comb chop cut sew broom tie fan

271

2. The reduplicated nouns The nouns that arise from the process of reduplication are of different types: true nominalizations, the result of the verbal activity (which in some cases corresponds with the internal θ-role Th), the external θ-role, and the internal θ-roles Th or Inst This sounds rather a chaotic situation, but we will show that in fact there is a fair degree of regularity. Before we start a discussion of the thematic roles adopted by these nouns, we will first discuss the few cases that appear to be true nominalizations, i.e. in the sense that the shooting by John of Bill involves a nominalization of the verb shoot. 2.1.

True Nominalizations

babari "tumult" might be construed as a plain nominalization of ban "to shout", together with laulau "folly, madness" from lau "crazy, insane". Both these we have assigned the argument structure ( )Th. Related would appear to be kosokoso "a cough" where we require to note that this may apply to a single instance of coughing. 2.2.

Result nouns

The following cases may be interpreted as resultative formations:

272

Norval S. Smith bronbron fonfon kofukofu koskosi kotkoti krabkrabu kraskrasi sakasaka ta(i)tai

2.3.

"burnt food-crust'h" "beating" "cuff' Μ "curses" "slice" "scrapings" "rash" "dregs" "bundle"

theme (with specific import)

theme theme theme (with specific import)

External arguments

The only cases of reduplications referring to an external argument are firstly, the semantically specialized cases of insects djompodjompo "grasshopper" from djompo "jump", and freifrei "0/ from frei "to fly", and secondly moimoi "finery" from moi "nice, beautiful", which clearly must be given a H e m e interpretation. 2.4.

Internal Theme

Apart from the four cases of result nouns mentioned above that are susceptible to an interpretation as theme, we have the following case: njanjan

'food'

This is a verb with the argument structure (Th)Ag. 2.5.

Instrument nouns gritgriti kankan kapkapu "krabbokrabbo" nanai sisibi titei tjatjari wawai

grater comb machete rake needle broom rope yoke, pad for freight fan

Deverbal nominalizations in Sranan: a search for regularities

273

3. Morphosyntactic regularities Now we are in a position to re-examine the list of formations in order to see whether there are any regularities pertaining to the meaning of the deverbal noun, and the argument structure of the original verb. The first thing that strikes the eye is that there is a clear relation between the verbs we have classified as (ΊΊι, Inst)Ag and the nouns that we have classified as Instrumentals. All eight of these verbs in our list form an Instrumental noun. Two verbs that we have not suggested are strictly subcategorized for an Inst - tjari "to cany" and krabu "to scratch" in one of its occurrences, also form an Instrumental noun. Note however that the notion "cany" does not necessarily imply the presence of a separate Instrument in the same way as verbs like "rake" and "sieve" do. Further, the actual sense of the Instrument noun is much less precisely constrained in this case. The case of "scratch" is slightly different, but we have assumed above that bodily actions do not involve an inherent instrument. This judgement is of course tentative. One criticism that could be made here is that we have assumed that titei "rope" is derived from the word tai "tie". This requires some clarification. In fact tei is not attested as a Sranan variant. However, it does occur in Ndjuka, a Bush Negro language, which linguistic and historical factors require us to consider as a development of early 18th century Sranan. The fact that tei occurs in Ndjuka guarantees it as an early variant of Sranan tai. A fairly regular relation also exists between verbs with the argument structure (Th)Ag, and the formation of what we have referred to as Result Nouns. Of the ten verbs with this argument structure, seven form Result nouns. In addition tai which we have classified as an Instrumental verb - forms a Result noun, in addition to also forming an Instrument Noun. Note that three of the monovalent verbs have reduplications that can be considered to be some form of nominalization, while in the other three cases, the reduplication refers to the external argument role. All in all a good result, considering the unproductive nature of the process in the small number of occurring cases.

274 4.

Norval S. Smith Phonology

At first sight we appear to be confronted with a fairly chaotic situation, with at least three classes of reduplication as follows: - full reduplication: e.g. sakasaka, laulau - reduced reduplication with closed syllable in initial element: e.g. gritgriti - reduced reduplication with open syllable in initial element: e.g. sisibi However, a brief glimpse at the list in (1) demonstrates that all the forms evidenced in the 18th century source consulted admit of the fully reduplicated form. During the last two hundred years we have evidence for reductions of the second and third type. 4.1.

Closed Syllable Base griti kapu kosi krabu krasi seki taki

4.2. a.

18th century

kosikosi krabokrabo krasikrasi takitaki

20th century

gritigriti fazpkapu Arakosi

grc'fgriti

ftabari ji'sibi tjaritjari/f/atjari moimoi/momoi nanai ietei wawai

fcabari i/sibi f/atjari moimoi nanzx ft'tei wowai

Aöjkosi krabkiabu krasikrasi/Ararkrasi kraslaasi sekseki takXaüd taktaSd

Open Syllable ban sibi tjari

baribari sibisibi/nsibi tjaritjari

b. nai tei wai 4.3.

19th century

nainai fetei waiwai

Discussion

With the exception of the form sisibi, we can analyse the two types of reduction appearing in reduplications in terms of the phonological structure of the base form. Closed syllable reduction occurs in forms with a CVCV structure where the second C is an obstruent. Where this second C is a sonorant, or is absent altogether - i.e.

Deverbal nominalkations in Sranan: a search for regularities

275

in C W base structure - we get open syllable reduction. Note that we appear not to have intermediate structures like *tjartjari, although, at least in modern Sranan, these are permitted structures. Compare words like aria, sarid etc. with r-final syllables. Closed syllable reduction only occurs where the second V is a high vowel - i or u. This would seem to indicate that it is related to a similar loss of high vowels internally in compounds. Compare the following cases: banktt-notu beri-pe branti-miri lousn-gotro It must be noted that the deletion of high vowels appears to be somewhat intermittent, at least as far as the dictionaries are concerned. Nevertheless we will assume it to be the same phenomenon as Closed Syllable Reduction. Note that examples such as beripe are suggestive of an earlier date for Open Syllable Reduction than Closed Syllable Reduction, otherwise we might expect that forms such as *tjartjari would occur. This is confirmed by the fact that OSR is evidenced in the 18th century source, while CSR is not. As it turns out we will not suggest a scenario where any real significance is attached to the first of these two facts. OSR appears also to affect only i and u - in cases like wai the diphthong may of course only end in i or u - although in the cases where a whole syllable disappears we only have final i. The identity of environment for the operation of OSR to that required for CSR may indicate that this was in fact basically the same process, occurring under differing phonotactic restrictions. We could hypothesize that the two phenomena do not overlap in time - that they appear to do so may be an artefact of the dictionary-producing activities of the 18th and 19th century lexicographers of Sranan. The difference between OSR and CSR could then purely be a reflection of the shift from a purely open syllabic structure in earlier Sranan, as against a new set of phonotactic conditions in later Sranan, whereby closed syllables are also allowed. Note that there is evidence for this anyway, completely independent of the phenomena we are dealing with here. This evidence, which we will not go into further here, derives from the reflexes of liquid clusters in the Surinam Creoles. We refer the reader to the discussion in Smith (1987). In other words, it is conceivable that we had a single reduction rule operative in Sranan in the 18th and 19th centuries. In the 18th century Sranan was basically

276

Norval S. Smith

restricted to an open syllable structure, with one exception: final nasals were allowed. Informally stated, the reduction rule appears to be statable as follows: preceding a stem, delete a stem-final high vowel preceded - directly or indirectly by another vowel within the same stem. In CVCV structures this has the effect that the second C is now stem-final - CV£-CVCV. However, this C would not be syllabifiable given the requirement that syllables be open, and would therefore not be realized. Following the change to the range of permitted structures, the above reduction rule would operate precisely as above, with the one difference that the C referred to above would now be syllabifiable, as closed syllables are now allowed. In the light of this we would have to absence of structures like *tjartjari as fortuitous. This hypothesis must remain tentative - we have only a small amount of evidence bearing on it so far. It would also appear to be the case that these reduplicated structures became susceptible to reduction - of whatever type - before compounds became a reduction environment. Unless the evidence for reduction was concealed here by a greater influence from a "standard" picture of the word. As the sources make clear "standardization" in the orthography played a significant role during this period (Smith 1987). In present-day Sranan, reduction is of even wider scope, operating between words. However, the precise conditions of application remain to be studied, as in fact does most of Sranan phonology.

5.

Conclusion

In conclusion we will sum up what we have done within the scope of this very brief article, which must be regarded rather as a set of pointers for future research rather than as any kind of definitive study of Sranan reduplication. We have examined two aspects of reduplication in Sranan - morphosyntactic and phonological. As far as morphosyntax is concerned we have studied the question of possible regularities holding between the argument structure of the verbs forming the basis for the reduplication process and the type of nominal resulting from the reduplication. Within the limited number of cases available it did seem that a considerable degree of regularity could be established. In terms of the phonology the interest lay in reductions visible in a number of cases, whereby the first element lost one or two final segments. These reductions could be grouped in two types, the first of which appeared to be older. The hypothesis was advanced that in fact the reductions involved the same process, and that it was the principles governing Sranan syllabification that had changed.

Deverbal nominalizations in Sranan: a search for regularities

277

In other words the phonogical phenomena are also less chaotic than they appear at first sight.

REFERENCES Alleyne, M.C. 1987. Studies in Saramaccan Language Structure. Caribbean Culture Studies 2, Amsterdam, University of Amsterdam. Schumann, C.L. 1778. Saramaccan Deutsches Wörter-Buch, Bambey, ms. Smith, N.S.H. 1987. The Genesis of the Creole Languages of Surinam, University of Amsterdam dissertation. Wullschlägel, H.R. 1856. Deutsch-Negerenglisches Wörterbuch. Nebst ein Anhang Negerengfische Sprüchwörter enthaltend. Lobau, T.U. Duroldt.

Aktionsart, aspect and duration adverbials

Co Vet Department of Romance Languages State University, Groningen

0.

Introduction*

This paper deals with two aspectual phenomena in French from an interpretative point of view. Section 1 deals with sentences in which transitional or non-transitional Aktionsart is combined with a duration adverbial, and with the different readings which these combinations occasion. Section 2 deals with the semantic effects which can be observed in sentences in the Imperfective Past of French when they are combined with a duration adverbial. It will be shown that the syntactic representations as proposed in Vet (1986) cannot always serve as input for the interpretation rules and that the grammar has to be provided with rules which, in some well-defined cases, give rise to partially interpreted syntactic structures which represent the semantic structure in a more satisfactory way.

1. Aktionsart and duration adverbials It will be assumed that the parser is capable of identifying the tense and the predication of a sentence and of representing them in the format proposed in

280

Co Vet

Vet (1986). For sentence (1) this yields the Syntactic Representation (SR) of (2): (1)

Marie remplit la bouteille 'Marie filled the bottle'

(2)

SP e,: [remplir^ (dlx,: Marie (x,))^ (dlx,: bouteille (x,))^,] (e,)

In (2) the tense operator SP (Pass6 Simple 'Simple Past') binds a variable (e() which stands for a time-space region. The predication (the part between the square brackets) expresses a 'situational concept' which is predicated of e,. In other words the predication describes what is going on on e,. By applying the predication to the time-space region e, the latter becomes a 'conceptualized time-space region' (see Bartsch 1986, 1989) or 'situation'. The term 'situation' is here used as a general term for 'event', 'state', etc. The nature of the predication determines the first temporal category I will deal with here, viz. Aktionsart. It is generally accepted that the category of the verb plays an important role in determining the Aktionsart of a sentence. In Vet (1980) it was argued that the main distinction is that between transitional and non-transitional verbs. The verb of sentence (1), remplir 'fill', belongs to the class of transitional verbs. The predications in which this type of verb is used describe a transition from a state A to a state Β (the resulting state). For (1) this resulting state is reached when the bottle is full. When such a resulting state is reached the situation cannot be continued. Non-transitional verbs do not lead to a resulting state. For example: (3)

Marie chercha la bouteille 'Marie looked for the bottle'

I will assume that the information about the transitional or non-transitional character of the verb is given in its lexical entry. For remplir and chercher the (partial) lexical entries are given in (4) and (5): (4)

remplir^ {\s)Ag (x2)Go meaning: CAUSE (x;) [T (-• rempli (x2)), (rempli (x2))]

where Vtr stands for 'transitional verb' and Τ for 'transition'; remplir 'fill' describes a transition from a situation in which some entity x2 is not full (-> rempli 'not filled') to a situation in which it is (rempli 'filled'). The lexical entry for the non-transitional verb chercher 'look for' is:

Aktionsart, Aspect and Duration Adverbials (5)

281

chercherK (x,)Ag (x2)Co meaning: try (x;) [find (x;) (x2)]

It is well-known from the literature that the presence or absence of the feature transitional determines the choice of the duration adverbial. Compare the following examples (in the SP): (6) a. b.

(7) a. b.

Marie remplit la bouteille en dix minutes 'Marie filled the bottle within ten minutes' #Marie remplit la bouteille pendant dix minutes 'Marie filled the bottle for ten minutes' Marie chercha la bouteille pendant dix minutes 'Marie looked for the bottle for ten minutes' #Marie chercha la bouteille en dix minutes 'Marie looked for the bottle within ten minutes'

Transitional verbs combine with en adverbials, non-transitional with pendant adverbials. For (6b) and (7b) special interpretation devices are needed (which is indicated by the symbol '#'). (6b) can be interpreted as referring to a set of filling events (frequentative reading); the adverbial indicates the duration of the set as a whole (this interpretation becomes more plausible if it is clear from the context that the bottle was to be filled many times by different persons). (7b) can only be interpretated as 'it took ten minutes before Mary started looking for the bottle'. The presence in (7b) of the adverbial en dix minutes Svithin ten minutes' forces the hearer to interpret the sentence as referring to a transitional situation, a transition from some preparatory event to the beginning of the looking-for activity/ For (6a) I propose the following SR (with a simplified notation for the terms):2 (8)

SP e,: [remplirKo. (Marie)^ (la bouteille)Go](el) : en dix minutes (e,·)

For (6b) it is as in (9): (9)

SP e,: [remplir^ (Marie)^ (la bouteille)^] (e,): pendant dix minutes (e,)

The interpretation rules can apply directly to the SR of (8), but not to that of (9) because sentence (6b) does not refer to one event, but to a set of events. I assume, therefore, that the grammar possesses interpretation rules which apply to SRs that contain a transitional verb combined with a pendant adverbial. These rules produce an 'Interpreted Syntactic Structure' (ISR), which may be

282

Co Vet

regarded as a first step towards a discourse interpretation ä la Kamp (see Kamp 1981). For (9) the ISR is as in (10): (10) SP Ε,: {e: [remplirKlr (Marie)^ (la bouteille^Ke)}: pendant dix minutes (E,) (10) has to be read as follows: there is a set E, of situations e in which Marie fills the bottle; the set E, has a duration of ten minutes. The SR of (7a) is as in (11): (11) SP e,: [chercher,, (Marie)^ (la bouteille)Go](e,): pendant dix minutes (e,) Since the verb chercher 'look for' does not belong to the transitional verbs no special interpretation rules are needed. The SR of (7b) is as in (12): (12) SP e(: [chercher^ (Marie)^ (la bouteille^Ke,): en dix minutes (e,·) The SR of (12) cannot be directly used as input for a discourse representation because it would give rise to an interpretation in which there is a looking-for event which has a duration of ten minutes. As we have seen, (7b) has to be interpreted as referring to a preparatory event which leads to the looking-for event. I will use the notation PREP(chercher) to refer to that preparatory event. The ISR for (7b) is then as follows: (13) SP e,: [PREP(chercher)^ (Marie)^ (la bouteille^Ke,): en dix minutes (e,) Meaning: Τ (-> chercher (x,) (x,)), (chercher (x,)(x,)) Note that PREP(chercher) is a transitional predicate which describes a transition from a non-looking-for situation to a looking-for situation. The reinterpretation of the verb chercher is caused by the presence in (7b) of the en adverbial, which combines only with transitional verbs. In the class of the transitional verbs a subclass of verbs has to be distinguished which does not behave in the same way as 'normal' transitional verbs. This is the class of momentary transitional verbs. The difference is that when they are combined with an en adverbial, this adverbial does not indicate the duration of the event referred to by the verb, but the preparatory process leading to it. For example: (14)

Marie atteignit le sommet en dix minutes 'Marie reached the summit within ten minutes'

Aktionsart, Aspect and Duration Adverbials

283

Obviously, the reaching of the summit itself does not last ten minutes. I propose the following interpretation. The lexical entry of atteindre 'reach' is as in (15) (15) a t t e i n d r e V t r + m (x,)^ (x2)Go meaning: CAUSE (x;) (Tm (-> atteint (x2)), (atteint (x2)) (15) tells us that atteindre belongs to the class of momentary transitional verbs (Vtr+m) and that it describes a momentary transition (Tm) between a state in which x2 was not reached and a state in which x2 was reached. The SR of (14) is as in (16): (16) SP e,: [atteindre V t r + m (Marie)^ (le sommet)Co)](e,): en dix minutes (e,) This syntactic structure has to be 'adapted' before it can be used as input for the interpretation rules. As for (7b) there has to be postulated a preparatory event (referred to by 'PREP(atteindre)'), which precedes the reaching event itself: (17) SP e;: [PREP(atteindreK, (Marie)^ (le sommet)Co](e;): en dix minutes (e;) meaning: CAUSE (x,) [T atteint (x,)), (e,)] Where e, is the spatial-temporal region on which the reaching event itself is realized: e(: [atteindreKff.+m (Marie (x,))^ (le sommet (x,))^,] (e,) The event in (17)) described by PREP (atteindre) is itself a non-momentary transition between a state in which the top is not reached and a momentary event in which Marie reaches the top. The ISR of (17) conforms to the characterization proposed by Vlach (1981: 288f) for the Vendlerian achievements such as reach the summit. He regards this type of situation as constituting the end point of an accomplishment. In (17) e, is the endpoint of Cj. The reinterpretation of the momentary verb atteindre as a non-momentary verb (PREP(atteindre)) is caused by the duration adverbial. In the next section I will examine the effects of the aspect carried by the Imperfective Past of French when it is combined with a duration adverbial.

284

Co Vet

2. The aspect of the Imparfait Sentences in the Imparfait 'Imperfective Past' (IP), do not refer to the same type of event as do sentences in the SP. Unlike SP sentence (1) its IP counterpart of (18) (18) Marie remplissait la bouteille 'Marie filled/was filling the bottle' does not permit of the conclusion that at any moment the bottle was full. Thus, whereas the SP sentence (1) expresses the situational concept as it is given in the lexicon, the use of the IP changes this concept, in (18) from transitional into non-transitional. The non-transitionality of (18) is a consequence of the fact that IP sentences usually do not refer to a maximal situation. For (18) this means that at the time to which the sentence refers Marie's filling of the bottle may be continued and, consequently, that the filling event has not been completed (and perhaps never was). For sentences in which the verb is non-transitional as in (19): (19) Marie cherchait la bouteille 'Marie looked for/was looking for the bottle' the effect of the IP is less clearly perceptible, but, just as (18), example (19) does not exclude that the looking-for situation had some continuation after the time at which (19) was true. The SP sentence (3) excludes this possibility (here Marie can only start looking for the bottle a second time, but in that case there is a new looking-for event, which is not presented as the continuation of the former event). In both cases the IP refers to non-maximal (or 'open') situations, i.e. situations without their boundaries (cf. also Bartsch, 1986, 1989). This means that IP sentences do not refer to the same type of situation as the corresponding SP sentences. If we adopt for (18) the SR of (20): (20) IP e(: [remplir^ (Marie (x,))^ (la bouteille (x,))] (e,) the ISR could be as in (21) (provisional representation): (21) IP e,: [NONMAX remplir^ (Marie (x,))^ (la bouteille (x;))] (e,) Where NONMAX is an (aspectual) operator which operates on the predicate, transforming it, in this case, from transitional into non-transitional.

Aktionsart, Aspect and Duration Adverbials

285

An interesting property of the IP sentences, probably related to their nonmaximal character, is that the situations they describe are non-countable. For example (22): (22) #Marie remplissait (IP) trois fois la bouteille 'Marie filled the bottle three times' cannot be interpreted as referring to three filling events, but has to be understood (with some difficulty) as referring to a repetition of complex situations each of which consists of three filling events. The corresponding SP sentence (22') refers to exactly three filling events: (22') Marie remplit (SP) trois fois la bouteille 'Marie filled the bottle three times' If IP sentences are combined with an en adverbial as in: (23) #Marie remplissait (IP) la bouteille en dix minutes 'Marie filled (=used to fill) the bottle within ten minutes' the only possible interpretation is again a frequentative one. Sentence (23) refers to a set of situations, each of which is a maximal event with a duration of ten minutes. The adverbial en dix minutes Svithin ten minutes' converts the non-count expression into a count one just as water can be used in count expressions such as a glass of water. Since the IP of (23) does not change the maximal character of the predication the IP cannot be regarded as a predicate operator (NONMAX) and, consequently, the representation of (21) has to be rejected. What the IP sentences of (18) and (23) have in common, however, is that they both express a kind of partitivity, comparable respectively with du sable 'sand' and des livres 'books' in: (24) II y avait du sable sur le tapis 'There was sand on the carpet' (25) II y avait des livres sur le tapis 'There were books on the carpet' Expressions such as du sable 'sand' and des livres 'books' have very little identificatory force and cannot be used in contexts where a referring expression is required (cf. de Smet, 1985), compare:

286

Co Vet

(26)

Le sable/Quelque sable/*Du sable 6tait sur le tapis 'The sand/Some sand/Sand was on the carpet'

(27)

Les livres/Quelques livres/*Des livres livres dtaient sur le tapis 'The books/Some books/Books were on the carpet'

The partitive article du/des distinguishes itself in this respect from other weak quantifiers such as quelque(s) 'some', plusieim 'several', etc., which can be used in sentences such as (26) and (27). In these sentences quelque(s) is interpreted in the same way as some in some of the boys, for example where the boys denotes a specified quantity of boys. In (24) and (25) the expressions du sable and des livres do not 'cut out' some part of a quantity of sand or books the spatial boundaries of which are defined by the context (see also Dik, 1989: 150f). In the same way, the use of the IP in (18) and (23) does not presuppose that there exists a complete event (or a well-defined set of situations) of which the (set o f ) situation(s) described by the IP sentence is a part. In order to refer to the referential weakness of expressions such as du sable and des livres as well as that of IP sentences I will introduce a quantifier, P, which expresses the weakest form of quantification. Expressions such as du sable and des livres can be represented as in (28b) and (29b): (28)

a.

du sable

b.

Px,: [sable^J (x,)

(29)

a.

des livres

b.

Ppx,: [livreNc] (x,)

where Nm: mass noun; p: plural; Nc: count noun. For (18), repeated here as (30a) the SR is as in (30b), the ISR as in (30c): (30)

a.

Marie remplissait la bouteille

b. c.

'Marie filled/was filling the bottle' IP e,: [remplirKe. (Marie)^ (la bouteille) Go ](e,) IP Pe ( : [remplirK (Marie)^ (la bouteille^Xe,)

The ISR of (30c) corresponds to (28b). Note that in (30c) the feature tr is absent. The quantifier Ρ only combines with verbs (or rather predications) which refer to non-transitional situations. Transitional predications are incompatible with the quantifier Ρ since they refer, by definition, to situations which possess boundaries, or at least a final boundary. In this respect transitional predications are comparable with count nouns, which have clearly defined spatial boundaries.

Aktionsart, Aspect and Duration Adverbials

287

For (23), repeated here as (31a), the SR is represented in (31b) and its ISR in (31c): (31) a. b. c.

Marie remplissait (IP) la bouteille en dix minutes 'Marie filled ( = used to fill) the bottle within ten minutes' IP e(: [remplirKo. (Marie)^ (la bouteille)Go](eI): en dix minutes (e,) IP PE,: {e: [remplirKo. (Marie)^ (la bouteille)^] (e)]: en dix minutes (e)} (E,)

I assume again that the grammar possesses rules that apply to SRs such as (31b), i.e. to SRs which contain an IP operator, a predication with a transitional verb, and an en adverbial. The output of these rule are ISRs like (31c). (31c) has to be read as follows: there is a set (E,) consisting of transitional situations (e) each of which has a duration of ten minutes. The quantifier Ρ indicates that the set E, has a very low referential profile since it is presented without its boundaries (in any case without the first and the last situations of the set). In this respect (31a) is comparable with the expression des livres "books' in (29b). This term des livres refers to a set of books with very vague contours, but the entities of the set themselves (the books) possess well defined boundaries. In (31a) the adverbial en dix minutes "within ten minutes' defines the boundary of each situation e of the set E,. In both (29) and (31) the quantifier Ρ combines only with count expressions if they are plural (i.e. refer to a set of entities or situations). The weak referential force of expressions such as (28a), (29a) and (31a) explains why they need a context which provides the referential setting in which they can be interpreted. In (24) and (25) this setting is provided by the locative adverbial sur le tapis 'on the carpet'. IP sentences also need some (temporal) anchor with the help of which the situations they refer to can be localized in time. A temporal adverbial or a SP sentence in the preceding discourse can function as such a temporal anchor: (32) a. b.

A huit heures Marie remplissait la bouteille 'At eight o'clock Marie filled/was filling the bottle' Pierre entra. Marie remplissait la bouteille. 'Pierre entered. Marie filled/was filling the bottle'

If the locative or temporal anchors are absent the expressions containing the quantifier Ρ cannot be interpreted.

288 3.

Co Vet Final remarks

At first sight the set of sentences for which ISRs, and thus special rules, are needed may seem arbitrary, but this is not the case. The sentences for which I have postulated special interpretation rules have in common that they combine elements which are semantically incompatible and which for their interpretation often need specific contexts. This is the case, for example, when a sentence with transitional Aktionsart, referring to a situation which has already clearcut temporal boundaries by itself, is combined with a pendant adverbial, which has the function to define the temporal boundaries of a situation (see (6b)). The same is true for the cases in which a sentence with non-transitional Aktionsart is combined with an en adverbial. This type of adverbial indicates the duration which is necessary to obtain some result, whereas the non-transitional Aktionsart refers to situations that do not lead to a result (see (7b)). The combination of the imperfective aspect of IP sentences with an en adverbial gives rise to the same kind of incompatibility (cf. (23)). This usually necessitates an interpretation in which there is a shift from the individual level (e) to the higher level of the set (E). The ISRs proposed in this paper account for this shift.

NOTES

* I am grateful to Dolf Hartveldt for his useful comments on an earlier version of this paper. 1. The combination of a non-transitional verb with an en adverbial is often regarded as unacceptable. However, this unacceptability is far from absolute as is shown by Berthonneau (1989: 721f), who gives examples such as: (i)

Injy a d6test6 son jockey en deux mois Injy hated his jockey in two months 'Injy started to hate his jockey after two months'

Note further that the nature of the determiner of terms (especially the partitive article) can also influence the compatibility of the pendant adverbial with a predication containing a transitional verb (cf. Verkuyl, 1972). I will not deal with this problem here. 2. In (8) the stacked structure (X : Y) is used for the representation of sentential satellites. This is not the standard notation in Functional Grammar (cf. Dik 1989: 201).

Aktionsart, Aspect and Duration Adverbials

289

REFERENCES Bartsch, R. 1986. On Aspectual Properties of Dutch and German Nominalizations. In: V.Lo Cascio & C.Vet (eds). Temporal Structure in Sentence and Discourse. Dordrecht: Foris, p.739. Bartsch, R. 1989. Tenses and Aspects in Discourse. Theoretical Linguistics 15,133-194. Berthonneau, A.-M. 1989. Composantes Linguistiques de la Rifirence Temporelle. Th&se d'Etat, Paris VII. Dik, S.C. 1978. Functional Grammar. Amsterdam: North-Holland. Dik, S.C. 1989. The Theory of Functional Grammar. Part I: The structure of the clause. Dordrecht: Foris. Kamp, Η. 1981. Evdnements, Reprdsentations Discursives et R6f6rence Temporelle. Langages 64, 39-64. Smet, L. de 1985. On the Semantics of the French NP des Ν'. In: H.Bennis & F.Beukema (eds). Linguistics in the Netherlands 1985. Dordrecht: Foris, p. 167-176. Verkuyl Η J . 1972. On the Compositional Nature of the Aspects. Dordrecht, Reidel. Vet, C. 1980. Temps, Aspects et Adverbes de Temps. Geneva: Droz. Vet, C. 1986. A Pragmatic Approach to Tense in Functional Grammar. Working Papers in Functional Grammar 16. Vlach, F. 1981. The Semantics of the Progressive. In: Ρ J . Tedeschi & A. Zaenen (eds). Syntax and Semantics 14: Tense and Aspect. San Diego, etc.: Academic Press, p. 271-292.

Some remarks on direct quotation in Kombai

Lourens J. de Vries YAPPER, Seksi Bahasa, Irian J ay a (The Reformed Development Foundation)

0.

Introduction

The topic of this article is the role which direct quotation-forms play in Kombai. a Papuan language of the Awyu-Ndumut family of southern Irian Jaya, Indonesia/ The first part of this article is descriptive (section 1 and 2), sketching the functions of direct quotation in the domains of reported speech and reported thought. Direct quotation plays an important and interesting role in the morphosyntax of speech act verbs, of intentional constructions and of emotion/thought expressions. The function of direct quotation in narrative discourse is to portray mental processes of participants as they react to the events narrated in the story. Reported speech and reported thought are expressed very similarly; they are formally non-distinct in some contexts. Thought is 'inner speech' from the Kombai point of view. Direct speech is so strongly preferred that we are forced to consider the possibility that Kombai does not distinguish between direct and indirect modes of reporting speech. The second part of the article (section 3) is concerned with two factors which may have favoured the use of direct quotation-forms in a wide range of contexts and construction types.

292

Lourens J. de Vries

The first factor has to do with the tendency in the verbal semantics of Papuan languages to break down an action into its component acts and express each one of these acts by a separate verb (Foley 1986:113). When an event has an implicit sub-act of speaking, Papuan languages tend to make this speech element explicit and this triggers the use of direct quotations in Kombai. The second factor favouring the use of direct quotations has to do with the notion of prefield complexity (Dik 1989: 350). I suggest that the use of direct quotations is one of a number of strategies in Kombai to reduce syntactic complexity in the pre-verbal area of the rigidly verb-final clause.

1. Reported speech Kombai has two interesting features in the domain of reported speech. First, an extreme preference for direct quotation; second, the obligatory use of direct quotations not only when assertions are reported but also when other types of speech acts are reported (praise, promise, warn, ask, etc.).2 As far as the preference for direct quotation is concerned, I have, in fact, thusfar not found any examples of indirect quotation. Since this may be due to incomplete data, it seems safer for the moment to use the phrase 'extreme preference for direct discourse'. But there is a distinct possibility that Kombai does not distinguish between direct and indirect discourse. Drabbe (1959:23,137) claims that Kaeti and Wambon do not have indirect forms of quotation. Kombai is related to Wambon: they all belong to the Awyu-Ndumut family. The absence of indirect discourse in Kombai would follow naturally from the tendency in Papuan languages to prefer direct discourse. Lower Grand Valley Dani, for example, has indirect forms but "Dani speakers show a distinct preference for direct quotation" (Bromley 1981: 271). Wilson (1988: 46) states that Yali, also a Dani-family language, does not have indirect discourse. The second interesting point is the obligatory use of direct quotations with all speech act verbs, not just with 'to say'. This feature is also found in other Papuan languages. Larson (1984) gives examples for Western Dani of the use of direct quotation with the following speech act types: testify, tell, invite, call, say, speak, command, consent, inquire. Consider the following Kombai examples: (1)

Khe luwa riga yademo- nane- ne he say.3sg.nonfut stone collect- imp.pl.- quote.sg 'He told them to collect stones.'

Some remarks on direct quotation in Kombai (2)

Nu yaboma- defene I be.illlsg.nonfut- conn- quote.sg 'He said he was ill.'

(3)

Nu wamedefene luwa I come.lsg.fut- quote.sg say.3sg.nonfut 'He promised to come.'

(4)

Nu luwofera- de gu I asklsg.nonfut you

293

gama ade- n- e join.SS eat- tr- 2sg.fut

khogona- n- ene allright- tr- conn- quote.sg Ί asked him to eat with us and he accepted the invitation.' The example (1) is a reported command, (2) a reported assertion, (3) a reported promise and in (4) an invitation and its response are reported. Clues as to the type of speech act can be found in both the quotation and in the quotation-margin. For example in (1), the imperative mood of the verb in the quotation is a clue to the command-nature of the reported speech act. Clues in the quotation-margin include the use of specific speech act verbs, often in combination with the generic speech act verb luwa- 'to say, to speak'. The generic speech act verb may also occur on its own (e.g. in the promise of (3)). Generic and specific speech act verbs may both precede and follow the quotation. Quite often they occur both in the right and in the left quotation-margin, as in (5): (5)

Khe ludima wamedefene luwa he promise.3sg.nonfut come.lsg.fut- quote.sg say.3sg.nonfut 'He promised to come.'

Speech act verbs and quotation-markers may also occur interspersed in the quotation, especially in long quotations (e.g. (11)). The content of the quotation in (6), together with the verb luwa- 'to say', identifies the reported speech act as an act of praise: (6)

Wa- luwa- no kho mofena yaferaboperf- say- 3pl.nonfut man that very.good'They praised that man.'

n- eneno tr- conn- quote.pl

Speech act verbs have their second arguments obligatorily expressed by direct quotations: direct quotations have been grammaticalised with speech act verbs. Thus John promised me bananas must be expressed as John promised (said): 7 shall

294

Lourens J. de Vries

give you bananas' and John praised me as John said: You are good'. Notice that the use of direct quotation in Kombai does not imply verbatim quotation (see note 2 also). When the encoding situation (see note 2) is very clear from the context, the quotation-marker may be the only element of the quotation-margin (e.g. (2)). When the quotation-margin is deliberately left unspecified and the encoding situation of the utterance cannot be inferred from the context, the quotationmarker becomes an evidential suffix. Compare: (7)

Ai khwui lefa pig theft do.3sg.nonfut 'He stole a pig.'

(8)

Ai khwui lefanene pig theft do.3sg.nonfut- quote.pl 'He stole a pig (hearsay).'

The addition of -nene in (8) indicates that the speaker wants to make clear that what he says is not based on his own observation. Evidential use of quotationmarkers also occurs in other Papuan languages, for example Asmat (C.L. Voorhoeve, p.c.) and Dani (Bromley 1981). The usual order of constituents in reported speech-constructions is as follows: (i) identification of quoted speaker(s), (ii) generic and/or specific speech act verb, (iii) quoted utterance(s), (iv) cliticized quotation-marker -ne (or one of its allomorphs), (v) generic and/or specific speech act verb. The only obligatory constituent is the quotation (the quoted utterance(s)). When the quotative nature of the quoted clause(s) is very clear from the context, the quotation-marker may be left out (e.g. (34) below) but this is rare. The quotation is neither deictically nor syntactically integrated in the clause of the quotation-margin. E.g. the quoted speaker of (5) is identified in the quotation by the first person form of the verb but in the quotation-margin by the third person pronoun Ahe 'he' showing that quotation and quotation-margin have different deictic centres. The quotation has the form of an independent clause and is syntactically non-embedded. The quotation-marker and the speech act verbs optionally preceding and following the quotation 'bracket off the quotation from the quotation-margin. Thus quotations are in certain respects 'extra-clausal constituents' (Dik 1989: 264) in Kombai. However, when the quotation is left out, the remaining clause-structure is incomplete since the quotation expresses the clausal second argument of the speech act verb in the quotation-margin. Although the quotation-clause is syntactically not a constituent of the margin-clause, it is semantically and pragmatically of course integrated in that clause.

Some remarks on direct quotation in Kombai

295

The quotation-marker {-ne} has the following allomorphs: -ne, -neno, -nene and -nera. The forms -ne and -neno express the opposition singular versus plural; -nene is a variant form of -neno in which vowel-harmony has taken place (De Vries 1989: 136). Compare: (9)

Khogona- n- ene allright- tr- conn- quote.sg 'He consented.'

(10) Khogona- n- eneno allright- tr- conn- quote.pl 'They consented.' The -n- in (9)/(10) is a transitional sound (De Vries 1989: 135). The connective -e will be discussed below. The allomorph -nera optionally replaces -ne and -neno but only sentencemedially; it consists of -ne and the Same Subject medial verb conjunction -ra (see De Vries 1989: 145). The forms -ne/-neno may occur both medially and finally. Since -ne is sometimes used in plural contexts (e.g. (11)) but -neno never in singular contexts, -ne is the basic allomorph (widest distribution). Consider (11): (11) Büwogo maru-fo luluwoma-no lüwobadi nage parent go.up-SS.seq warn3pl.nonfut child you(pl) kheiforü mofene deilonefa nage taboo.meat that if.you.eat you(pl) afokhononera luluwoma-no gwari stay.little.2pl.fut- quote warn3pl.nonfut snake mune- y- enera muno- tr- conn- quote

fira fuwof-one fira rodent fuwo- conn- quote.sg rodent

yami- y- ene mofena-n-a kheiforü- forii yami- tr- conn- quote.sg thattr- conn meat.taboo- taboo mene deilonefa nage topic if.you.eat you(pl) fe- rabogo- nemadü- y- ene neg- big.grow- inf.fut- neg.fut- tr- conn- quote.sg

296

Lourens J. de Vries 'The adults use to warn us, Children, if you eat forbidden meat, you will stay little, the Muno snake, the Fuwo rat, the Yami rat, if you eat those forbidden kinds of meat, you will not grow up.'

The allomorphs of -ne point to its verbal origin. In some other Awyu family languages the cognates of -ne function fully as verbs of speaking. The verb ne(secondary stem nende-) 'to say* occurs in Mandobo (Drabbe 1959: 23) in the right margin of quotations, with the connective -o optionally linking the quotation to the verb of speaking: (12) Wagarew-o ne- gen badconn say- 3sg.pres '"It is bad", he said.' (13) Mendanog- ο nende- ran come.imp.sg- conn say3sg.past 'He told him to come.' The Kombai verb luwa-' to say" has the role of generic verb of speaking which nehas in Mandobo and, most probably, Kombai -ne has developed into a reported speech marker from this verbal origin. The plural form -neno consists of -ne and the 2/3 pl.nonfut ending -no which now only signals plurality of quoted speakers. Similarly, the medial quote-marker -nera contains the medial verb conjunctionclitic -ra. The optional connective -o of Mandobo corresponds to the -e which optionally but usually precedes the quotation-markers in Kombai. Having discussed the role and some formal aspects of quotation in reported speech, we now turn to the function of direct quotation in reported thought.

2. Reported thought Direct quotation-forms play an important role in the domain of reported thought, too. Mental processes (intention, attitudes, emotions) are often portrayed as a form of speech, 'inner speech' a person directs to himself. Formally, the quotation is the same whether used in reported speech or reported thought. The difference between the two is optionally expressed in the quotationmargin by lexical means. It is true that the generic verb of speaking cannot be used in certain reported thought contexts but in other reported thought contexts this verb, luwa- 'to say, speak', is obligatory (see examples (16)/(17) below). The verbs rerama- 'to be glad' (ex. (27)) and kharabuma- 'to be astonished' are examples of

Some remarks on direct quotation in Kombai

297

lexical means in the quotation-margin indicating the type of mental process in reported thought-constructions. We may distinguish between intentional thoughts and other (non-intentional) thoughts when describing the role of direct quotation in reported thought. 2.1.

Intention

The use of quotative elements and forms of 'to say' in intentional contexts is a very common feature of Papuan languages (Foley 1986: 156-157). An important factor determining the use of quotations in intentional constructions is whether intender and intended are the same or not (Foley 1986: 156). In (14), for example, the intender {John) and the intended person {Mary) differ but in (15) they are the same: (14) John wants Mary to leave (15) John wants to leave According to Foley (1986), in (14) an act of communication between John and Mary is essentially involved, from the point of view of Papuan languages. Indeed, when intender and intended differ in Kombai, the use of direct quotation-forms is obligatory: (16) Camate luwa kho yademo- nane- ne head.of.district say.3sg.nonfut person meetimp.pl- quote.sg 'The district-head wants the people to come together.' (17) Guru ya luwa- no furuma khakhe-nane- neno teacher them say- 3pl.nonfut well listen- imp.pl- quote.pl 'The teachers want them to listen well.' The verb of speaking luwa- is used in this type of intentional construction. (16) and (17) can be used to report speech and to report intention (thought). When intender and intended are the same, the category of person is an important factor determining the form of intentional expressions. With first person intenders, Kombai uses intentional verb-forms, without quotations: (18) Nu okh ami I water drink. lsg.intent Ί want to drink water.'/'Let me drink water.'

298

Lourens J. de Vries

(19) Ai- fo go- lplintent 'We want to go.'/'Let us go.' However, quotation is used with first person intenders in one type of intentional construction in which motion-verbs are used and plural intentional forms of the verb in the direct quotation: (20) Nu me- lara ai galemo-fonera I come-stand-SS.and pig buylpl.intent-quote wa- me- de perf- come-lsg.nonfut Ί have come to buy the pig.' Motion-verbs both precede and follow the direct quotation in (20). With second/third person intenders, Kombai uses the first person intentional verb-forms in combination with direct quotation: (21) Ai- fonene go- lplintent- quote.pl 'They want to go.' (22) Yarimo kho fera- fene garden go.SS see- lsg.intent-conn- quote.sg 'He wants to see his garden.' I have not found, thusfar, a lexical item 'to want' but there are lexical items for negative intention ('to not want something'). Examples are the adjectives bukhe and bidoge, both meaning 'reluctant' and the verbs lome- and biduma- 'to refuse'. All these items, whether verbal or adjectival, take nominal infinitive complements: (23) Khaη- ο biduma-no go.inf.nonfut- tr- conn refuse- 3pl.nonfut 'They do not want to go.' (24) Nu khaη- ο bukhe I go.inf.nonfut- tr- conn reluctant Ί do not want to go.' The connective -o links the infinitival complements to the verbal head of the clause in (23) and to the adjectival head of the clause in (24).

299

Some remarks on direct quotation in Kombai

Although it is possible to express negative intention without quotation-forms (as in (23)/(24)), negative intention may also be expressed with direct quotation and that is the most frequent way of expressing negative intention. Thus, of (25) and (26), the preferred option seems to be (26): (25) Obate miη- ο lome- ge medicine drink.inf.nonfut- tr- conn refuse- 3sg.nonfut 'He refuses to drink the medicine' (26) Obate miη- ο lome- defene medicine drink.inf.nonfut- tr- conn refuse- lsg.nonfut- conn- quote 'He refuses to drink the medicine' We may conclude that, when intender and intended are the same, the use of direct quotation is not obligatory in all cases. With first person intenders, quotation is not used (cf. (18)/(19)), except for purposive constructions with motion-verbs (cf. (20)). For second/third person intenders, quotation is obligatory with positive intention but with negative intention lexical means are available that do not involve direct quotation (cf. (23)/(24)), although even in that case the form using direct quotation (ex. (26)) seems to be the preferred option. These facts point to the tendency in Kombai to use direct quotation not only when intender and intended differ but also when they are the same. This is because there is, from the Kombai point of view, an act of communication involved both when intender and intended differ and when they are the same, although in this last case perhaps in a weaker sense: in that case the intender communicates with himself, in 'inner speech'. 2.2.

Other types of reported thought

Because direct quotations form an obligatory part of the morpho-syntax of certain types of intentional constructions, they have been, to a certain extent, grammaticalised in the domain of intentionality. In the domain of reported speech (see 1), they have been grammaticalised completely. By grammaticalisation of direct quotation I mean that it must be used in certain construction-types independent of discourse-factors. When other (non-intentional) types of thought are portrayed with direct quotation, discourse-factors crucially determine this use. Consider the following examples: (27) Khe rerama burii he be.glad.3sg.nonfut clanterritory 'He is glad to go to his own place.'

ai- fene go- lsg.fut- quote.sg

300

Lourens J. de Vries

(28) Ya imimo kharabuma- no they all be.astonished- 3pl.nonfut

khe fenemora he how

makhe- y- ene do.3sg.nonfut- Qtr- conn- quote.sg 'They were all astonished because of the things he did.' (29) Yafo- finä wa- khumoleineno their- thought perf- die.3sg.nonfut-quote.pl 'They think he is dead.' The various kinds of mental processes are indicated in the quotation-margins of (27) - (29). The noun finä in (29) means "breath", but also 'thought' and 'desire'; the verb finäge- 'to think', 'to long for' is derived with the verbalizing support-verb -khe from the noun finä (cf. De Vries 1989: 140-141). This verb can be used both as an experiential and as a non-experiential predicate. Compare: (30) Ya okh- ο finäge they water-conn desire- 3sg.nonfut 'They are thirsty." (31) Nagu momof- ο finäge we uncle- conn desire- 3sg.nonfut 'We long for our uncle.' (32) Kho mofene finä- genon- a besawa person that think-3pl.nonfut- tr- DS.and plane meda-nemadii come-inf.fut- neg "Those people think that the plane will not come.' The Subject person-number argeements in (30) and (31) indicate the experiential nature of these clauses: the verb does not agree with the Experiencer. Literally, (31) says: '(It) desires us of uncle'. In this way the experiencer is portrayed as not controlling the mental process (cf. De Vries 1989: 140, 141). The verb finäge- is used as a non-experiential verb in (32), in which it agrees with kho mofene 'those people' who are portrayed as controllers of the mental process. Finäge- and fibima- 'to think' have their second arguments expressed in the final clause of a co-subordinate clause-linkage. Consider:

Some remarks on direct quotation in Kombai

301

(33) Ya fibima- non- a khegabükhe they think- 3pl.nonfut- tr- DS.and he angry "They think he is angry.' The second argument-clause khe gabiikhe 'he is angry" in (33) is syntactically independent, not embedded in the preceding medial clause containing fibima- 'to think'. The switch-reference (Different Subject following clause) co-subordinator -a links the two clauses. We shall return below to this use of 'co-subordination'. The choice between using direct quotation (ex. (27) - (29)) or co-subordination (ex. (32) - (33)) for expressing non-intentional thought seems to be determined by pragmatic factors. I do not pretend to have understood these factors fully and I limit myself to sketching one type of context which triggers the use of direct quotation. We find this context when in narrative discourse the mental processes occurring in participants are portrayed as mental reactions to the event of the narrative. Consider (34): (34) Wa,fenemora ya na- lu fe- khakhudo oh how they my-word neg- listen.inf.non-fut- neg 'He was perplexed that they did not listen to him.' The example (34) is the last sentence of a short narrative in which the main participant tries to teach in his native village but the village people, to his surprise and disappointment, reject him. Then (34) follows, without any quotation-margin (even the quotation-marker is absent), to portray the thoughts of surprise and disappointment occurring in the main participant. For the Kombai listeners to this story, it is very clear from the preceding context what the encoding situation for (34) is, i.e. whose 'inner speech' is quoted. The function of the direct quotation in the story is clear to the Kombai listeners because they are familiar with the narrative convention of portraying mental reactions by direct quotations. Κ there occurs a verb in the quotation-margin in such narrative contexts, it is usually a verb of thinking or feeling, often with emotional colour, like kharabuma'to be astonished' in (28). Literally, (28) says: 'They were astonished, "how does he do it?".' The same type of narrative context of emotional participant-reaction triggers direct quotation in Western Dani (Larson 1984) and there we also find exclamative interjections (like wa in (34)) initially in the quotation and verbs of emotion in the quotation-margin.

302 3.

Lourens J. de Vries Factors favouring the use of direct quotation

It should be clear by now that direct quotation plays an important role in the syntax and discourse of Kombai. In this section, I shall, very tentatively, sketch two factors which may be relevant to the frequency of direct quotation and its use in a wide range of construction-types. 3.1.

Specificity of event-description

Foley (1986: 113-128) describes a tendency in the verbal semantics of Papuan languages which may be relevant to the use of direct quotation in Kombai. This tendency is to be very specific in the description of events, by breaking down actions into their component acts and express each one of these components by a separate verb (Foley 1986: 113). An extreme but illuminating example is from Kalam: (35) Yad am mon pk d ap ay- pyn I go wood hit hold come put-perf- lsg Ί fetched firewood.' For Kalam, to be sure, (35) is not at all extreme: to leave out any of the verbs would make the description seem incomplete to Kalam speakers. This tendency to spell out the sub-events of an event in detail manifests itself also in Kombai. For example, the event of 'going' is usually described as follows: (36)

Khe ragane maru lara kha he rise.SS go.up.SS stand-SS.and go.3sg.nonfut 'He went away.'

Some other examples: (37) Dunoro fali me food cariy.SS come.3sg.nonfut 'He brought food.' (38) Foro moja ma- none carry.SS go-down.SS come-imp.pl 'Bring it (down)!' The events of bringing and taking are always described as a combination of 'carry' and 'come'/'go'. Often (as in (38)) there is also a motion-verb specifying the direction.

Some remarks on direct quotation in Kombai

303

Now when an event has a speech-component, Papuan languages tend to make explicit this speech-element by a verb of speaking and in Kombai this triggers the use of direct quotations (see 2.). Since, from the Kombai point of view, mental events like 'think' and "want' also imply speech ('inner speech'), the explicitation of this element of speech triggers direct quotation in those domains, too (cf. (16)/(17)). 3.2.

Prefield compexity in a rigid verb-final language

The second factor which may help to explain the importance of direct quotation in Kombai is a syntactic one and has to do with constituent ordering tendencies in language. Kombai is a rigidly verb-final language with the basic functional pattern S Ο V. Now subordinate clauses are complex constituents syntactically (Dik 1978: 204212) and several supposedly universal tendencies conspire to remove these complex constituents from the pre-verbal area of the clause (prefield) and position them after the verb (in the postfield). For example, the Prefield Complexity Principle formulated by Dik (1989: 350): "The Prefield is universally less hospitable to complex material than the Postfield; Prefield languages may thus be expected to possess strategies for relieving the Prefield of excessive complexity." Another tendency is that "Other things being equal, constituents prefer to be placed in order of increasing complexity" (Dik 1989: 351). Thus subordinate clauses in a rigid verb-final language like Kombai are subjected to conflicting ordering pressures. On the one hand the language tends to adhere to its basic functional pattern with the predicate (whether verbal or nominal) in final position. On the other hand, the aforementioned tendencies put pressure on the pre-verbal complex constituent to move to the area after the verb. There is no problem for topical subordinate clauses in Kombai. They go to the special clause-initial position (PI in the terminology of Dik (1989)), reserved, inter alia, for constituents with topical functions. This position is insensitive to complexity-pressures (Dik 1978: 204). Focus-clauses do not go to PI in Kombai; their focality is expressed by the Focus-clitic -a on the relational headnoun (cf. De Vries 1989: 185, 220). But not all subordinate clauses are topical in Kombai (De Vries 1989: 220). An important class of non-topical subordinate clauses are the second argument-clauses of verbs of speaking, thinking and wanting and of perception-verbs ('to see', 'to hear'). It follows from the ordering theory of Dik (1989) that Kombai should possess strategies to solve the conflict of verb-finality and pre-verbal complexity. One solution would be to reduce the complexity of pre-verbal subordinate clauses, instead of postponing them.

304

Lourens J. de Vries

As in the majority of rigid verb-final languages (Dik 1989: 378), the Kombai strategy of reduction is to give pre-verbal embedded clauses a nominalized form. Consider (39) and (40): (39) Nu khaη- ο biduma-de I go.inf.nonfut- tr- conn refuse- lsg.nonfiit Ί do not want to go.' (40) Ya uwa rakhuma- no they money pay3pl.nonfut 'They forgot to pay.'

ro enuma- no thing forget- 3pl.nonfut

The examples (39) and (40) represent the two forms that embedded pre-verbal clauses take in Kombai. The first (such as in (39)) is with a nominal infinitive linked by the prefield-centre connective -o to its head, the verb of the superordinate clause. The second and more usual form is to express the subordinate clause as a relative clause modifier of a relational headnoun, the two most common headnouns being ro 'thing' and rumu 'person' (cf. De Vries 1989: 216-222). Thus (cf. (40)) embedded clauses take the form of noun phrases containing relative clauses. Noun phrases are less complex than fully verbal embedded clauses (Dik 1989: 351). However, nominalized non-topical clauses (such as (39) and (40)) are relatively infrequent and this is because Kombai prefers two other strategies, direct quotation and co-subordination. These two strategies give the complex constituent an independent syntactic status so that it is no longer a constituent of the superordinate clause. Let us limit our attention to clausal second arguments ('object clauses'). The great majority of these clauses are expressed by direct quotation and/or cosubordination. Verbs of speaking (see section 1) use direct quotation for their clausal second arguments: (41)

Khe luwa riga yademo- nane- ne he say.3sg.nonfut stone collect- imp.pl- quote.sg 'He told them to collect stones.'

In section 2 we discussed the fact that direct quotation-clauses are neither deictically nor syntactically integrated in the clause of the quotation-margin. Therefore, e.g. in (41) it is no problem that the clausal second argument of luwa occurs after the verb. The speech act verbs preceding and following the quotation merely "bracket off that constituent from the quotation-margin. Direct quotations allow for highly complex material (cf. (11) above). We have seen (section 3.) that

Some remarks on direct quotation in Kombai

305

direct quotation is by no means restricted to reported speech but also occurs with 'think' and 'want'. Co-subordination also gives the complex constituent (pre-verbal embedded clause) an independent syntactic status, but in a different way. The term 'co-subordination' (Olson 1981) refers to the combination of dependence and nonembeddedness which characterizes the relationship between medial and final clauses in Papuan languages. (See Foley and Van Valin 1984: 241-242, 256-263 for detailed treatment of co-subordination.) Co-subordination is used for the second argument-clauses of many verbs, for example for perception-verbs: (42) Nu fera- defa khebo- me I see- lsg.nonfut- DS.and he dur-come.3sg.nonfut Ί saw him come.' The example (42) could be paraphrased as: Ί saw and he was coming', to give the reader a rough idea of co-subordinate clause-linkage, but the purely coordinate rendering with 'and' fails to indicate the type of dependency between the two clauses of (42) although it succeeds in rendering the non-embedded nature of the relationship. The dependency of medial clauses on the following final clause (independent clause) has to do with tense-mood and person-number interpretation of the verb of the medial clause (these depending in varying degrees on the verb of the final clause) and with switch-reference relations between medial and final clauses. The Different Subject form feradefa Ί saw and Different Subject following' cannot stand on its own. For reasons of available space we cannot elaborate here on co-subordination and we shall limit our attention to the aspect of complexity. The combination of non-embeddedness and dependency makes co-subordination ideally suited for solving the problem of verb-finality and pre-verbal complexity. The complex constituent (e.g. the clausal second argument khe borne in (42)) occurs after the verb of perception but its independent syntactic status as a final clause prevents violation of the SOV pattern: it is not a constituent of the preceding medial clause. With some verbs, both the quotation and the co-subordination strategy may be used (compare e.g. (29) and (32)) but usually either direct quotation (e.g. ludima'to promise') or co-subordination (e.g. fera-'to see') is used for the second argument. There are a few cases in the data where certain verbs of speaking (luwa- 'to say, speak', ja- 'to call') have their second argument-clauses expressed by cosubordination and not by direct quotation. The verbs of speaking have direct causative meanings which they do not have when direct quotation is used:

306

Lourens J. de Vries

(43) Nu bo- luwa- nan- a ai ü-de me dur-say- 3pl.nonfut- tr- DS.and pig kill- lsg.nonfut 'They forced me to kill the pig.' (44) Luwa riga yademo- nane- ne say.3sg.nonfut stone collect- imp.pl- quote.sg 'He told (them) to collect stones.' (45) Ya ja- nan- a nagu me- defo they call-3pl.nonfut- tr- DS.and we come-lpl.nonfut 'We came on their invitation.' The examples (43) and (45) show the causative meaning appearing in cosubordinate environments with certain verbs of speaking. Luwa in (44) is noncausative and non-factive: the speaker of (44) does not suppose that the collecting indeed took place whereas in (43) and (45) the factuality of the killing and the coming is presupposed. 3 More research is needed to establish whether the use of co-subordination with verbs of speaking always implies causative meanings. But it is clear that cosubordination and direct quotation cannot be simply viewed as interchangeable strategies. They share, however, one important feature: they give complex constituents of clausal form a non-embedded status and in doing so help to solve the conflict between pre-verbal (prefield) complexity and verb-finality. 4.

Summary

Direct quotation plays a role in the following domains. The second arguments of speech act verbs are expressed by direct quotationclauses (e.g., (l)-(6)). In a few cases, these arguments are expressed by cosubordination (cf. (43), (45)) but in those cases the verbs of speaking have causative meanings. Reported speech and reported thought are handled similarly in Kombai: direct quotation is used in both domains. When intentional thoughts are reported, a crucial factor is whether intender and intended are the same or not. When they differ, direct quotation is used obligatorily (cf. (16)-(17)). When they are the same, the situation is more complicated. With first person intenders, quotation is not used (see (18)-(19)) except for constructions with motion-verbs (see (20)). For second/third person intenders, quotation is obligatory with positive intention but optional with negative intention. In the domain of non-intentional thought, direct quotation is used in narrative discourse to portray mental processes of participants (e.g. (34)).

Some remarks on direct quotation in Kombai

307

The syntactic relation between direct quotation and quotation-margin shows considerable independence for the quotation-clause. The non-embeddedness of the direct quotation may be one of the reasons why direct quotation is such a preferred strategy in Kombai, In a rigid verb-final language, the direct quotation strategy makes it possible , for complex material to be expressed after the verb, or rather independent of the position of the verb, since the complex constituent is not in construction with the verb. Another reason for the importance of direct quotation in Kombai may be the tendency in Papuan languages Uo make explicit the sub-acts of actions. When the speech-element in events is described separately by a verb of speaking, this implies in Kombai the use of direct quotation. From the Kombai point of view, mental processes also have a speech-component ('inner speech') and this extends the range of constructions in which direct quotation is used. Since the use of quotative forms in intentional constructions is very widespread in Papuan languages, it could well be that the Kombai tendency to express reported speech and reported thought very similarly is a manifestation of a genferal Papuan typological characteristic.

NOTES 1. Kombai is spoken by about 4000 people living in the isolated swampy rainforests of the UpperDigul area of southern Irian Jaya. I have lived in this area since the end of 1982. Kombai fieldwork started in february 1986. The data on quotation were collected during the second half of 1989. The Awyu-Ndumut family (cf. Voorhoeve 1975:27) belongs to the Trans-New Guinea Phylum of Papuan languages (cf. Silzer and Heikkinen 1984). For a sketch of Kombai grammar: see De Vries 1989. Information on Kombai culture can be found in Venema 1989 and De Vries 1987. 2. Following Dooley (1989), we shall define (in)direct discourse in terms of deictic orientation and not in terms of verbatim versus adapted modes of reporting. Direct quotation-forms have all their deictic elements oriented immediately in Kombai, i.e. oriented towards the utterance-situation of the reported and not of the reporting speech act. The deictic definiton works better for Kombai since, in the absence of the distinction direct versus indirect discourse, the use of direct forms does not imply anything about the attemps of the speaker to report verbatim. Also following Dooley (1989: 45), we shall use the term 'quotation' for the quoted utterance(s), and 'quotation-margin' for the constituents in the reported speech constructions which describe the encoding situation, the cicumstances under which the quoted utterance(s) was (were) uttered; in John said to Harry: You are ill' the quotation-margin is John said to Harry and You are ill' the quotation. The present tense and the second person pronoun you show the immediate deictic orientation of the quotation. 3. Notice that the factive presuppositions of (43) and (45) follow from the causative meaning of the verbs of speaking in co-subordinate environments and not form the co-subordinate pattern in se. The example (33) shows that co-subordinate structures expressing clausal second arguments may carry non-factive presuppositions.

308

Lourens J. de Vries

REFERENCES Bromley, H.M. 1981. A Grammar of Lower Grand Valley Dani. Pacific Linguistic Series C 63. Canberra: Australian National University. Dik, S.C. 1978. Functional Grammar. Amsterdam: North-Holland. Dik, S.C. 1989. The theory of Functional Grammar. Part I: The structure of the clause. Dordrecht: Foris. Dooley, R A . 1989. Suggestions for the Field Linguist Regarding Quotations. Notes on Linguistics, no 44, 34-50. Drabbe, Father P. 1959. Kaeti en Wambon. Twee Awju-dialecten. The Hague: Nijhoff. Verhandelingen van het Koninklijk Instituut voor Taal-, Land- en Volkenkunde. Foley, W. A. 1986. The Papuan languages of New Guinea. Cambridge: Cambridge University Press. Foley, W A . and R.D. van Valin Jr. 1984. Functional syntax and universal grammer. Cambridge: Cambridge University Press. Larson, G.F. 1984. The use of direct quotation in Western Dani Bible translation. Unpubl. paper. Olson, M. 1981. Barai clause junctures: towards a functional theory of interclausal relations. Unpubl. Ph.D. thesis Australian National University. Silzer, Ρ J . and H.L. Heikkinen. 1984. Index of Irian Jays languages. Irian, Bulletin of Irian Jaya, vol. ΧΠ. Venema, H. 1989. Sagogrub festival. Irian, Bulletin of Irian Jaya, vol. XVII, 38-63. Voorhoeve, C.L. 1975. Languages of Irian Jaya: checklist, preliminary classification, language maps, wordlists. Pacific Linguistics. Series B, n. 31. Canberra: Australian National University. Vries, L J . de, 1987. Kombai Kinship terminology. Irian, Bulletin of Irian Jaya, vol XV, 105-118. Vries, L J . de, 1989. Studies in Wambon and Kombai. Aspects of two Papuan languages of Irian Jaya. Unpubl. Ph.D. dissertation University of Amsterdam. Wilson, J.D. 1988. Scripture in an oral culture: the Yali of Irian Jaya. Unpubl. MA thesis, University of Edinburgh, Faculty of Divinity.

Some remarks on direct quotation in Kombai ABBREVIATIONS 1 2 3

first

conn DS dur fut imp inf intent nonfut perf pi pres Q quote seq sg supp SS tr

person second person third person

connective the following clause has a Different Subject durative future imperative infinitive intent non-future perfective plural present question-marker quotation-marker sequence singular verbalizing support-verb the following clause has the Same Subject transitional sound

309

INDEX

Language Index Alamblak 214ff. Amele 215 Araucanian 9ff. Asmat 218f. Aymara 9 Bahasa Indonesia 243ff. Dutch 211., 77f. English 93ff. French 279ff. German 232ff. Golin 216 Hua 214ff. Hungarian 82ff., 241f. Kalam 302

Kobon 213ff. Kombai 291ff. Latin 19£f., 195£f. Lillooet 47ff. Mandobo 296 Quechua 2ff., 151ff. Romance 195ff. Salish 57ff. Shuswap 57 Spanish 107ff. Sranan 265ff. Thompson 57f. Usan 213ff.

Subject Index abstract nouns 204ff. adjuncts 48 adpositional term predicates 93ff. adpositions 195ff. adverb 211f., 254 adverbial 214, 279ff. adverbial particle 215 affected object 54f. Agent 268ff. Aktionsart 279ff. applied translation studies 188 argument structure 268ff. aspect 52, 77f.( 85f., 284 assertives 113ff. auxiliary operators 80ff. candidate answers 137ff. case 196ff. categorial identity thesis 230ff. categoriality 89 causativity 50 Change 60 circumpositions 233ff.

cislocative 154f., 157 classifiers 245ff. cleft constructions 103f. closed question 133 cognitive functions 36ff. coherence 29ff. complement clauses 3f., 9, 19ff. conditional 112f. consciousness 158f. context model 40 control 51ff., 108ff. conversation analysis 135, 165 conversational closings 175 copula support 94f. co-subordination 304ff. cultural'functions 42ff. dative 201f. , derivation 48f., 98ff. derivational· morphology 151f. descriptive approach to translation 185 determiner phrase 244ff. direct speech Iff., 19ff., 291ff.

312

Index

directives 58, 62, 119ff. discourse 27ff. discourse grammar 28 discourse representation 282 discourse structure 165ff. discourse unit 135 drift 200f. duration adverbials 279ff. ecology of language 71f. ethnomethodology 165 evidential 8 Exclamation 20 expression rules 73ff. Explanation 30f. extra-clausal constituents 14ff., 21, 294 farewell rituals 175f. functional coherence 30ff. functional discourse analysis 27ff. Functional Grammar 13ff., 37f., 59f., 78ff., 93ff., 226f. functional head 230 functional pragmatics 33ff. functional text analysis 29ff. functional text semantics 29ff. genitive 255f. Government and Binding theory 243ff. head-to-head movement 251 hedged performatives 114f. horizontal functionality 42f. iconic space 116 illocutionary conditions lllff., 119 illocutionary satellites 18f. Imparfait 284ff imperative 87 impersonal constructions 86f. implied object 48 independent linguistic entity 13f. indirect speech Iff., 19ff., 291ff. infinitive 74ff. inner speech 4, 291ff. Instrument 203ff., 268ff. intentionality 297ff. interpreted syntactic structure 281f. interviewing technique 133ff., 165ff. intransitivizers 48f. intransitivity 48ff. layers 16ff„ 207, 212, 223ff. leading question 134 linguistic translation 184 literary translation 184 locative noun phrase 156

locative prepositions 96f. macro speech act 34f. Manner 203ff. manner adverbs 215ff. maxim of quantity 117, 125 measure words 245ff. merging of cases 199 metaphorical space 113 mitigation 116ff. model structures 38ff. models 37 modification 248 Momentariness 282f. morphological change 200 negation 223 nominal predicates lOOff. nominalization 265ff. non-restrictive relative clause 260 nonsubject constituents 5f., 10 nouns 69f. open question 133 opening rituals 170f. operators 79ff., 212, 284f. orientation 239 origin of language 65ff. particle 212 participle 75f. partitivity 285f. perspective 161 phonological change 198 placement rules 80 politeness 107ff. positive politeness 107ff. possible world 112, 129 postpositions 233ff. pragmatic functions 14ff. predicate 212ff. predicate formation 79ff., 219 predicate satellites 226 prefield 303ff. prepositions 229ff. - see also adpositions prescriptive approach to translation primary operators 81 proposition satellites 225f. proximity 239 quantification 286 quantifiers 260ff. questions 133ff. question particles 21ff. quotation Iff., 19ff., 291ff.

Index racist talk 33 reduplication 265ff. referential coherence 30 relational transitivizers 55f. restrictive relative clause 252f. rhetorical functions 35ff. satellites 17ff., lOOff., 207, 212ff. scripts 41 semantic functions 37f. sentence 66f. situational concept 280 social functions 42f. spatial adverbs 214f. spatial reference 161 speech act sequences 33ff. specifier 230 stranding 238 style disjuncts 18 superstructures 32 switch reference 8f. temporal adverbs 212ff. temporal reference 161. tense operator 280 term predicate formation 93ff. text linguistics 28 Theme constituents 15f. Theta-roles 265ff. time dimension 68 time reference 76f. Transition 280,288 transitivity 51ff. transitivizers 49ff. translation research 184ff. translation studies 184ff. translation teaching 186ff. translation theory 187f., 189 translocative ISSf. trigger rules 80 turn taking 135ff. verbs 69f. - of communication 2ff., 50f., 53, 292ff. - of knowledge 3f. - of movement 152f., 154f. - perception 3f., 50f., 53 - of reflection 3f., 296ff. - weather verbs 153, 157f. - 'to say" 4f., 11 verb class 152ff. verb serialization 218f., 224 vertical functionality 42f.

313 word order 232ff., 303ff. X-bar theory 229ff.