Studies on Reduplication 9783110911466, 9783110181197

For several reasons, mostly inherent to the different developments of generative grammar, an increasing number of public

268 46 20MB

English Pages 649 [652] Year 2005

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Editorial Preface
Introduction
Reduplication: Form, function and distribution
From repetition to reduplication in Riau Indonesian
Morphological Doubling Theory: Evidence for morphological doubling in reduplication
The Emergence of the Marked: Tone in some African reduplicative systems
Reduplication and consonant mutation in the Northern Atlantic languages
Wrong side reduplication is epiphenomenal: Evidence from Yoruba
Non-adjacency in reduplication
Enhancing contrast in reduplication
Phrasal reduplication and dual description
Reduplication in Modern Hindi and the theory of reduplication
On the grammaticalization of verbal reduplication in Japanese
Reduplicative allomorphy and language prehistory in Uto-Aztecan
Reduplication in Tupi-Guarani languages: Going into opposite directions
On the borderline of reduplication: Gemination and other consonant doubling in Arabic morphology
Syntactic reduplication in Arabic
Reduplication in the Vedic verb: Indo-European inheritance, analogy and iconicity
Reduplication in child language
Reduplication before age two
Acquisition of reduplication in Turkish
Reduplication in Pidgins and Creoles
Less is more: Evidence from diminutive reduplication in Caribbean Creole languages
Intensity and diminution triggered by reduplicating morphology: Janus-faced iconicity
Backward and sideward reduplication in German Sign Language
A reanalysis of reduplication in American Sign Language
List of key words
List of languages
List of contributors
Recommend Papers

Studies on Reduplication
 9783110911466, 9783110181197

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

Studies on Reduplication

w DE

G

Empirical Approaches to Language Typology 28

Editors Georg Bossong Bernard Comrie Yaron Matras

M o u t o n de Gruyter Berlin · New York

Studies on Reduplication

edited by Bernhard Hurch with Editorial Assistance of Veronika Mattes

Mouton de Gruyter Berlin · New York

M o u t o n de Gruyter (formerly M o u t o n , T h e Hague) is a Division of Walter de Gruyter G m b H & Go. K G , Berlin.

© Printed on acid-free paper which falls within the guidelines of the A N S I to ensure p e r m a n e n c e and durability.

Library of Congress Cataloging-in-Publication

Data

Studies on reduplication / edited by B e r n h a r d H u r c h . p. cm. — (Empirical approaches to language typology ; 28) Includes bibliographical references and index. I S B N 3-1 1 - 0 1 8 1 1 9 - 3 (cloth : alk. paper) 1. G r a m m a r , C o m p a r a t i v e and general — Reduplication. I. H u r c h , Bernhard. II. Series. P245.S855 2005 415 - dc22 2005002331

ISBN 3 11 0 1 8 1 1 9 3 Bibliographic

information

published

by Die Deutsche

Bibliothek

Die Deutsche B i b l i o t h e k lists this publication m the Deutsche Nationalbibliografie; detailed bibliographic data is available m the Internet at < h t t p : / / d n b . d d b . d e > .

© Copyright 2 0 0 5 by Walter de Gruyter G m b H & C o . K G , D - 1 0 7 8 5 Berlin. All rights reserved, including those of translation into foreign languages. N o part of this b o o k may be reproduced or transmitted in any form or by any means, electronic or m e c h a n i c a l , including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany.

Editorial Preface

The present volume originated with the Graz Conference on Reduplication, which took place in early November 2002 organized by the Institute of Linguistics of the University of Graz and hosted by the European Center for Modern Languages (European Council). The meeting was sponsored by the Federal Ministry of Science, the University of Graz, the province of Styria, and the City of Graz. The construction of a typologically oriented database on reduplication is the topic of a research project at the Graz Linguistics Institute, with financial support of the Austrian National Bank. Both - the database and the present volume - are responses to a widely perceived lack of empirical breadth and of a notable repetition of examples in the discussion of reduplication. From the very beginning all contributors were invited to strengthen their theoretical claims with broad substantive evidence. And the evidence presented at the conference ranges from historical comparative studies to typological issues, from language acquisition to pidgins and Creoles, from the discussion of possible origins of reduplication to its functional and semantic properties, from formal preferences to new descriptive proposals, from sign languages to the formulation of theoretical principles; moreover, they embrace a wide diversity of languages and language families of the world. The authors emerge from different theoretical frameworks without an a priori agreement on theoretical assumptions. Stressing the substantive evidence of a phenomenon always means also stressing its diversity. In Humboldt's terms: in order to achieve the mediation between phenomenology and typology, we must integrate the study [des] Besonderen im Allgemeinen through the study [des]Allgemeinen im Besonderen. And we hope that this volume is a step in this direction. Our deepest gratitude goes to all contributors and discussants of the conference, to the present authors, to the large number of reviewers who have seriously contributed to the improvement of the volume, to the European Center for Modern Languages, to the sponsoring institutions, and especially to Bernard Comrie as editor of the EALT-series for his dedicated help and support in the editorial work. Graz, June 2004

Bernhard Hurch & Veronika Mattes

Contents

Editorial Preface

ν

Introduction Bernhard Hurch

1

Reduplication: Form, function and distribution Carl Rubino

11

From repetition to reduplication in Riau Indonesian David Gil

31

Morphological Doubling Theory: Evidence for morphological doubling in reduplication Sharon Inkelas

65

The Emergence of the Marked: Tone in some African reduplicative systems Laura J. Downing

89

Reduplication and consonant mutation in the Northern Atlantic languages Fiona Mc Laughlin

Ill

Wrong side reduplication is epiphenomenal: Evidence from Yoruba Nicole Nelson

135

Non-adjacency in reduplication Patricia A. Shaw

161

Enhancing contrast in reduplication Suzanne Urbanczyk

211

Phrasal reduplication and dual description Elinor Keane

239

viii

Contents

Reduplication in Modern Hindi and the theory of reduplication Rajendra Singh

263

On the grammaticalization of verbal reduplication in Japanese Reijirou Shibasaki

283

Reduplicative allomorphy and language prehistory in Uto-Aztecan Jason D. Haugen

315

Reduplication in Tupi-Guarani languages: Going into opposite directions Frangoise Rose

351

On the borderline of reduplication: Gemination and other consonant doubling in Arabic morphology Dina El Zarka

369

Syntactic reduplication in Arabic Utz Maas Reduplication in the Vedic verb: Indo-European inheritance, analogy and iconicity Leonid Kulikov

395

431

Reduplication in child language Wolfgang U. Dressier, Katarzyna Dziubalska-Koiaczyk, Natalia Gagarina, Marianne Kilani-Schoch

455

Reduplication before age two Marie Leroy andAliyah Morgenstern

475

Acquisition of reduplication in Turkish Hatice Sofu

493

Reduplication in Pidgins and Creoles Peter Bakker and Mikael Parkvall

511

Less is more: Evidence from diminutive reduplication in Caribbean Creole languages Silvia Kouwenberg and Darlene LaCharite

533

Intensity and diminution triggered by reduplicating morphology: Janus-faced iconicity Werner Abraham

547

Contents

ix

Backward and sideward reduplication in German Sign Language Roland Pfau and Markus Steinbach

569

A reanalysis of reduplication in American Sign Language Ronnie B. Wilbur

595

List of keywords

625

List of languages

631

List of contributors

637

Introduction Bernhard Hurch

Reduplication denotes a morphological procedure by which the inflectional and/or derivational formatives used to signal a specific category are directly derivable from the phonological/prosodic structure of the uninfected or underived simplex form. There are some fine scholarly treatments of general principles of reduplication in a broad overview, ranging with a long break from August Friedrich Pott's (1862) Über Doppelung to Ronnie Wilbur's (1973) seminal dissertation and the treatment of reduplication in the framework of the Stanford Universals Project by Edith Moravcsik (1978). Reduplication itself is not as rare a phenomenon in the languages of the world as one might judge from the Indo-European perspective, nor is it formally restricted to the types we find in Greek or Latin. Carl Rubino opens the volume with the paper Reduplication: Form, function and distribution. In a very general way he describes and quotes the different formal types of reduplication found in the languages of the world; he gives an overview of the semantic and grammatical functions reduplication may serve, and finally briefly summarizes aspects of distribution. Diachronic hypotheses on reduplication are rare. They are mostly based on indirect evidence from the distribution of types over grammatical functions (e.g., Bybee et al. 1994) and propose some type of origin and development by analogy to grammaticalization (Niepokuj 1997). We only have very few instances of direct observation of the rise of reduplication, which, in addition, point to somewhat different origins of the procedure. 1 David Gil sketches in From repetition to reduplication in Riau Indonesian a series of criteria for distinguishing between repetitive and reduplicative constructions. One of the strong diachronic hypotheses, namely that the origin of partial reduplication is (always) to be found in full reduplication, has some strong support in David Gil's study, which detects a gradual passage between these two iterative constructions, from repetition to reduplication. The question of whether reduplication would better be described in phonological or in morphological terms is a vaexata quaestio of the study of reduplication. It has been closely linked to the more general question of the

2

Bernhard Hurch

status of morphology in grammar. The obvious grammatical function of reduplication is counterbalanced by the fact that most reduplication can be described entirely in phonological terms, as it very much resembles phonological rules with a structural description and a structural change, mostly describable with reference to phonological units. Wilbur's (1973) seminal dissertation, well received in other respects, has been widely ignored for stressing precisely this misleading character of reduplication, while the generative tradition of the last two decades strongly favored approaches based on phonological descriptions. 2 In this sense Sharon Inkelas' paper Morphological Doubling Theory: Evidence for morphological doubling in reduplication represents a radical way of re-thinking the issue. The major idea behind Morphological Doubling Theory is that reduplication is seen as a morphological construction whose individual parts are morphosemantically identical. Both partial and total reduplication follow the same principle, with the first one being subject to further truncation. Strong arguments in favor of morphosemantic identity adduced in this paper include affix reduplication, suppletive allomorphy and other morphotactic discrepancies between the simplex (base) and the reduplicant structure. Moreover, Inkelas clearly separates morphological and phonological elements involved in reduplication and accordingly also argues that backcopying works through analogy to phonological duplication. The relation of the reduplicant to the base is, in addition to the already mentioned diachronic relevance, also important for statements regarding their asymmetry. This refers to the fact that the reduplicant, and in this respect it behaves like other affixes, usually has a simpler phonological and prosodic structure than the base (especially if it refers to a bound morphological category). Affix reduplications, for example, are generally expected to be structurally less marked. In the article The Emergence of the Marked: Tone in some African reduplicative systems Laura Downing describes tonal asymmetries which point in the opposite direction, as they exhibit more highly marked tonal structures on the reduplicant than on the base. Her explanation recalls the essentially autosegmental character of tone and explains tonal asymmetries in reduplication for those cases and languages in which other affixes may also carry dominant tone structures. Backcopying, as outlined by Wilbur (1973), is a process by which a phonological stem alteration occurs under the influence of a reduplicative element. This process itself poses a problem for most theories, as there is wide agreement about the precedence of base over reduplicant. Correspondence theories deal precisely with arranging formal identity between the two ele-

Introduction

3

ments of reduplication. Fiona Mc Laughlin discusses an interesting example of so-called backcopying in her contribution Reduplication and consonant mutation in the Northern Atlantic languages. She discusses the different shapes the same type of consonant mutation takes in Pulaar and Seereer-Siin, two related languages in the area and argues against phonological backcopying. She analyzes the apparent assimilatory force as the result of different application procedures of the noun class affix, which triggers the mutation. In Seereer-Siin both the reduplicant stem as well as the base stem undergo consonant mutation, and it looks as if the second was influenced by the first. Mc Laughlin prefers a morphological approach to reduplication, which postulates that the two stems are a priori formally independent of each other and also undergo consonant mutation independently in Seereer-Siin. Wrong side reduplication refers to the situation in which a last portion of a morphological unit (e.g., word) reduplicates to the left of the whole unit, that is, it gets prefixed, or, conversely, a first portion of a unit attaches to the right, that is, gets suffixed. Then the reduplicant is not adjacent to (that portion of) the base, it is said to be derived from. Obviously this should be a dispreferred type of reduplication, although it is not explicitly excluded by most theories. In Wrong side reduplication is epiphenomenal: Evidence from Yoruba Nicole Nelson aims at eliminating wrong side reduplication altogether by explaining the examples from the literature as being the result of grammatical operations, different in nature, which then, incidentally, give rise to the appearance of wrong side reduplication. This nonlocality may arise, for example, when the "wrong side" is the result of deletion processes, when other specific regularities interact (like a strong rhyme reduplication), etc. Patricia Shaw touches upon one of the most intriguing problems for formal theories on reduplication, namely the situation in which base and reduplicant are not adjacent to each other. The difficulties arise equally for derivational as for non-derivational theories. In Non-adjacency in reduplication she uses the full range of data from one of the families with maybe the most complex reduplication systems, Salish, to show that in order to adequately explain the facts we simply need - contrary to previous proposals within her framework - principled restrictions about the character of the base as pertaining to either a system of morphological or prosodic categories, and secondly a constrained dependency relation between those constraints dealing with direct mapping of segmental content and those responsible for locating the reduplicative element. She thus delivers a theory

4

Bernhard Hurch

which on the one hand predicts that adjacency between base and reduplicant reflects the less marked situation, and which, at the same time, offers a principled account for cases in which adjacency is disrupted. In Enhancing contrast in reduplication Suzanne Urbanczyk describes a systematic morphophonemic disambiguation tendency which frequently applies in languages that possess more than one reduplication procedure from which specific reduplication types are used for different grammatical functions. In order to facilitate the parsing of the grammatical differences there are formal changes - strategies for the enhancement of the contrast that comprise vowel deletions in the stem, fixed segmentism in restricted contexts, differing use of vocalic space, differing accentual behaviour, etc. As a subsequent step the author searches for parallels to the described tendencies among processes of enhancement of phonological contrasts in general and opens up a series of interesting questions that will have to be explored by future research. Phrasal reduplication and dual description is the title of the contribution by Elinor Keane. She discusses echo formation taken from data from the literature on Bengali and Kannada and partly from her own data on Hindi and Tamil. One of the main points she makes is that she is trying to outline a systematic differentiation of the constituents working as inputs for echo-reduplications. The hypothesis she puts forward is that the typological difference might be anchored in the genealogy of the languages under scrutiny, with Indo-Aryan languages showing prosodic and Dravidian morphosyntactic bases. The formal model Keane is developing, dual description, is based on assumptions from declarative phonology and integrates for the specific purpose at hand formal refinements such as default inheritance and default unification. The paper by Rajendra Singh Reduplication in Modern Hindi and the theory of reduplication deals with the same subcontinent but focuses from among the rich variety of reduplication in Hindi (cf. Abbi 1992) on those compounding full reduplicative procedures which give evidence for a lexico-semantic analysis of reduplication, thus approaching the above outlined Morphological Doubling Theory. Reijirou Shibasaki deals with diachronic aspects of verbal reduplication in Japanese. The title of his contribution On the grammaticalization of verbal reduplication in Japanese indicates the theoretical frame in which he locates his analysis. The basic claim is that verbal reduplications in Japanese have changed their original verbal properties from Old to Modern Japanese into adverbials, in the sense that they have changed from lexical verbs to adverbial constructions. Shibasaki additionally proposes that ver-

Introduction

5

bal reduplications have been strongly adverbialized in Modern Japanese. He tries to substantiate his claims with a quantitative analysis of Japanese texts from the past 1100 years. The second study with at least some diachronic orientation deals with Uto-Aztecan. Jason Haugen in his contribution entitled Reduplicative allomorphy and language prehistory in Uto-Aztecan tries to find out how much a comparative study of the different reduplicative types found in the modern Uto-Aztecan (UA) languages can contribute to a detailed understanding of the genetic relationship between the languages, especially between the northern and southern branches. Haugen proposes reconstructing the most invariant UA reduplication types (light syllable reduplication; disyllabic reduplication; mora affixation; marked heavy syllable reduplication) for Proto-UA, which are present - in isolation or in combination - in all the branches of UA, and in full combination only in Yaqui. Fran^oise Rose, Reduplication in Tupi-Guarani languages: Going into opposite directions, gives an overview of the complicated patterns, relations and developments of reduplication within that one language family, with special attention to her own field studies of Emerillon. These languages possess basically two reduplication types (mono- vs. disyllabic) with two different meanings (event-internal vs. event-external plurality). But whereas monosyllabic reduplication is final in the better known TupiGuarani languages, it copies the first stem syllable in Emerillon; in parallel to the monosyllabic pattern, the disyllabic pattern copies the last two syllables with some obvious preference for a leftward reduplication in most languages. Interestingly the monosyllabic pattern seems to be recessive diachronically, being replaced by the newly competing disyllabic option. Rose attributes the different reduplication patterns and directions within the family to different influences in the rather widespread areal distribution. The size of the reduplicant has been subject of much controversy. Dina El Zarka in her On the borderline of reduplication: Gemination and other consonant doubling in Arabic morphology discusses what usually is considered the minimum for fulfilling the requirement for reduplication, the doubling of a segment. The canonical structure she is discussing is CVCCVC in which the bold consonant is the result of reduplication. This picture of a simple consonant working as a reduplicant - following El Zarka - perfectly fits the overall picture of Arabic morphology. She excludes the purely phonological/prosodic character of this alternation, while what brings it closer to reduplication in the classical sense is semantics. These forms frequently denote augmentatives, intensives, iteratives, etc.

6

Bernhard Hurch

Moreover, she enters the diachronic discussion with the argument that these partial reduplications can never be traced back to full reduplication, the latter being the origin of partial reduplication in the analysis of various authors. Another quality of El Zarka's paper is that she draws attention to a set of data which up to now has been neglected and that she calls into the discussion arguments from very different varieties of Arabic and Semitic in general. Utz Maas in Syntactic reduplication in Arabic situates his reflections at the other end of the scale. After a long introduction to the history of reduplication research and the different understandings of reduplication and especially of its domain, he focuses on which has been termed in rhetorical studies paronomasia, i.e., syntactic reduplications. As a relatively productive form, even if stylistically restricted, he analyzes the masdar constructions, specifically the masdar reduplications of Maltese and Moroccan Arabic, and establishes a typological difference to Indo-European with respect to this point. Leonid Kulikov dedicates his study Reduplication in the Vedic verb: Indo-European inheritance, analogy and iconicity to this oldest attested Indo-Aryan language. Vedic has a rich system of at least five types of verbal reduplication. It is considered to have partly inherited and preserved the original Indo-European reduplicative system, and partly to have developed new types. There has often been postulated a specific correlation between reduplication and child language, reduplication being a characteristic trait at certain periods of language acquisition, even in the acquisition of languages which in the adult state do not regularly use reduplication as a grammatical means. In this context it also has been hypothesized that the occurence of early reduplication might be one of the origins of reduplicative mechanisms in adult language. In the present volume three papers deal with different aspects of reduplication in language acquisition. Wolfgang Dressler et al. {Reduplication in child language) discuss some of the general questions of the interrelation of child language and reduplication, specifically about plain and expressive morphology, about the role reduplication has for the acquisition of morphology, about specific types and characteristics of reduplication and their origin in child language. Marie Leroy and Aliyah Morgenstern limit themselves to early acquisition stages. In the article Reduplication before age two they detect the most different strategies for the use of reduplicative devices for pragmatic purposes, for pregrammatical functions, and in word-formation with referential value. They conclude that reduplication is used at the beginning as a motor phenomenon that helps to

Introduction

7

control speech mechanisms and slowly gets limited to communicative and then lexical purposes. They call reduplication a poly-functional strategy, which over time bridges the gap between a playful practice and the construction of linguistic skills. Hatice Sofu on the other hand studies the Acquisition of reduplication in Turkish. The acquisition of the pattern of full reduplication seems not to be very problematic, contrary to the pattern of partial reduplication, also called emphatic reduplication in Turkish, which expresses an intensification of adjectives and adverbs. Sofu attributes the problems children acquiring Turkish morphology have with this morphological type to basically two factors: partial reduplication is the only prefixing operation in Turkish which otherwise has a full agglutinating system of suffixation, and second to the complicated rules choosing C 2 in the prefixing reduplicant CiVC 2 -, where Ci but not C 2 depends directly on the base. In addition she briefly mentions the questionable productivity in general of partial reduplications in Turkish. One type of language which makes extensive use of reduplication is formed by pidgins and Creoles (Kouwenberg 2003). It is not entirely clear how strong a typological argument one can make from these languages, for example with respect to the discussion of the historical origin of reduplication, but still the variety which exists must be part of human cognitive systems. In their contribution Reduplication in pidgins and Creoles Peter Bakker and Mikael Parkvall give an overview of the full variety of reduplication types and their grammatical/semantic functions in pidgins and Creoles and make a series of observations, which contradict current assumptions on these language types. As historically young languages, Creoles show a relatively high proportion of full in relation to partial reduplications, but we can neither postulate a stronger iconic interpretation than in non-creoles, nor assume a higher number of strongly iconic instances of reduplication (e.g., plural, which is encoded more readily following the strategies of the substrates). In addition to a series of other claims the authors clearly show that, contrary to explicit statements in the literature, pidgins are poor in their use of reduplications. Interestingly the process of creolization, that is, the mechanism of expansion, re-vitalizes specific functions of the substrates and explains the higher rate of reduplications in Creoles, due to substrates. Silvia Kouwenberg and Darlene LaCharite search for a plausible way to semantically and cognitively connect meanings and functions of reduplication, which - within a theory of iconicity - seem at a first glance unre-

8

Bernhard Hurch

lated, if not contradictory. Their paper Less is more: Evidence from diminutive reduplication in Carribean Creole languages explores a way to link the apparently opposite meanings of plurality and diminution. Both continuity and dispersivity are instantiations of a concept of "more", where the first one approaches plurality in the classical sense, whereas the second possibly attenuates the intensity of the original meaning through the scattered presence of a given characteristic. It thus bridges the gap between plurality on the one hand, and diminution and approximation on the other. Werner Abraham (Intensity and diminution triggered by reduplicating morphology: Janus-faced iconicity) takes up - in a kind of comment-paper - the same line of interpretation from Kouwenberg & LaCharite and proposes a solution of these seemingly contradictory functions of reduplication in terms of intensional and extensional modification of the base. On the other hand Abraham expresses much stronger scepticism towards the concept of iconicity, but without clear reference to one of the established uses of the concept. In sign languages as well reduplication seems to be an operation which is amply used in different grammatical and expressive morphological functions. Roland Pfau and Markus Steinbach treat two of these aspects in their article Backward and sideward reduplication in German Sign Language: the role of backward reduplication for the formation of reciprocals and of sideward reduplication, where the second movement is displaced horizontally with respect to the first, for the formation of nominal plurals. These reduplication types must be integrated with the verb type (onehanded and two-handed agreeing verbs), whereas plain verbs do not reduplicate and the reciprocal form is not marked. Correspondingly the plural formation by reduplication in German Sign Language depends on phonological features. Reduplication takes place with nouns, which are signed in the neutral signing space; simple reduplication applies on nouns that are signed in relation to the midsagittal plane, whereas lateral signs require sideward reduplication. Again there is an unmarked group of nouns in plural formation, namely those with body-anchored signs and those with complex movements. Ronnie B. Wilbur in A reanalysis of reduplication in American Sign Language sketches a theory of the mapping between meaning and form in the morphology of ASL within a system of temporal and spatial primitives that build on the direct representation (expression) of argument structure. The distinction between repetition and reduplication can easily be stated in terms of the number of movements and especially through the direct iconic-

Introduction

9

ity of the number of repetitions; reduplications must have at least three repetitions of the lexical portion of the predicate, whereas repetitions are interpreted as having a single return movement. Traditional treatments of reduplication in ASL associate meaning with the velocity of the movement: slow reduplications express continuation, fast reduplications habitual or iterative forms. In nouns the arc movement signals plurality (collectives). In the remainder of the paper Wilbur delineates the basics of event structure she needs for sketching the verb types within the geometry and set theory she needs for a potentially universal theory of possibilities and limitations from which sign languages can recruit their expressive means. She further develops the uses ASL makes of these geometrical and set theoretical primitives and collocates them with reduplication. The mapping of these different combinations onto concrete production time results, among other things, in the representation of certain verbal aktionsarten as reduplicative form. The elongation of time within one event without endpoint and end state (e.g., through a curved line of the movement) expresses durativity, whereas the time structure between the events draws the difference between incessant, habitual, and iterative structures. A final part of the paper treats the limitations on the expression of distributivity, especially with its intrinsic relation to iterativity. The present collection is intended to be a state-of-the-art volume. It reflects a series of different positions, different problems, different perspectives on one broad problem - reduplication. And reduplication lends itself particularly well to perspectives opening to all directions of research and theoretical orientation: it touches on the question of the form-meaning relation and iconicity, of the interaction of phonology/prosody and morphology, plain and expressive morphology, of morphology and syntax, of categorial semantics, of grammaticalization, productivity, diachronic development; but also of a series of formal and functional principles like adjacency, sequentiality, contrast enhancement, markedness relations. These and a number of more specific arguments are treated in this volume, and we hope the reader will find it as enjoyable and stimulating as was the Graz Reduplication Conference, which was the origin of this volume.

10

Bernhard

Hurch

Notes 1. 2.

For a comprehensive discussion of this issue, cf. Hurch & Mattes (2004). Including Prosodic Morphology, where phonological (prosodic) issues are the main discussion points.

References Abbi, Anvita 1992 Reduplication in South Asian Languages. An Areal, Typological and Historical Study. New Delhi etc.: Allied Publishers. Bybee, Joan, Revere D. Perkins and William Pagliuca 1994 The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Hurch, Bernhard and Veronika Mattes 2004 Über die Entstehung von partieller Reduplikation. In Sprache und Natürlichkeit. Gedenkband für Willi Mayerthaler, Gertraud FenkOczlon and Christian Winkler (eds.),137-156. Tübingen: Narr. Kouwenberg, Silvia (ed.) 2003 Twice as Meaningful. Reduplication in Pidgins, Creoles and other Contact Languages (Westminster Creolistics Series 8). London: Battlebridge. Moravcsik, Edith 1978 Reduplicative constructions. In Universals of Human Language. Vol. 3, Word Structure, Joseph Greenberg (ed.), 297-334. Stanford: Stanford University Press. Niepokuj, Mary 1997 The Development of Verbal Reduplication in Indo-European. (Journal of Indoeuropean Studies Monograph 24), Washington: Institute for the Study of Man. Pott, August Friedrich 1862 Doppelung (Reduplikation, Gemination) als eines der wichtigsten Bildungsmittel der Sprache, beleuchtet aus Sprachen aller Welttheile. Lemgo & Detmold: Meyer. Wilbur, Ronnie 1973 The Phonology of Reduplication. Bloomington: Indiana University Linguistics Club.

Reduplication: Form, function and distribution Carl Rubino

The systematic repetition of phonological material within a word for semantic or grammatical purposes is known as reduplication, a widely used morphological device in a substantial number of the languages spanning the globe. This paper will provide an overview of the types of reduplicative constructions found in the languages of the world and the functions they portray. Finally, a subset of the world's languages, will be categorized as to whether or not they employ reduplicative constructions productively and illustrated in a world map'.

1.

Form

For purposes of the accompanying typological map, two types of reduplication are distinguished based on the size of the reduplicant: full vs. partial. Full reduplication is the repetition of an entire word, word stem (root with one or more affixes), or root, e.g. Tausug (Austronesian, Philippines) full word lexical reduplication dayang 'madam' vs. dayangdayang 'princess'; laway 'saliva' vs. laway-laway 'land snail', or full root reduplication, shown here with the verbalizing affixes mag- and -(h)un which do not participate in the reduplication: mag-bichara 'speak' vs. mag-bichara-bichara 'spread rumors, gossip'; mag-tabid 'twist' vs mag-tabid-tabid 'make cassava rope confection'; suga-hun 'be heated by sun' vs. suga-suga-hiin 'develop prickly heat rash' (Hassan et al 1994). Partial reduplication may come in a variety of forms, from simple consonant gemination or vowel lengthening to a nearly complete copy of a base. In Pangasinan (Austronesian, Philippines) various forms of reduplication are used to form plural nouns. (1.)

too 'man' > totöo CV- 'people'; amigo 'friend' > -CV- amimigo 'friends'; bdley 'town' > CVC- balbaley 'towns'; plato 'plate' > CjV- papldto 'plates'; manok 'chicken' CVCV- > manomanok 'chickens'; and dueg 'water buffalo' > Ce- dereweg /dedeueg/ 'water buffaloes'. (Rubino 2001a)

12

Carl Rubino

Ilocano (Austronesian, Philippines) employs a number of types of partial reduplication with various word classes, where the reduplicated material can be a partial root, simple root, a partial stem (bimorphemic entity), or a full word: (2.) Ilocano Reduplication Reduplicant Shape -C-

Use

Examples

Animate/kin plurals

CV-

Plural argument; Animate plurals General plurals; Imperfective aspect; Comparison

laldki 'male' > lalläki 'males' babai 'female' > babbdi 'females' ubing 'child' > ubbing 'children' na-lukmeg 'fat' > na-lulukmeg 'fat, distributive' ka-ili-άη 'townmate' > kakailiän 'townmates' kalding 'goat' > kalkalding 'goats' ag-bäsa 'read' > ag-basbäsa 'reading' dakkel 'big' > dakdakkel 'bigger' na-sam?it 'sweet' > na-samsam?it 'sweeter'

Lexical iterativity Mutuality Lexicalized items

ag-tilmon 'swallow' > ag-tilmotilmon 'swallow repeatedly' riipa 'face' > rupanriipa 'face to face' bänga 'pot' > bansabdnsa 'skull' tukdk 'frog' > tukaktükak 'wart'

CVC-

CVC(C)VCVC(C)VNFull

Partial Reduplication Across Morpheme Boundaries Reduplicant Shape CV-

CVC-

Affixes Involved Pa- causative; -inn- reciprocal ma- potentive pa-causative

Examples Ag-gi-pa-basol=zda VERB-PL-CAUS-blame=3p

'they are blaming each other' ma-turog 'sleep' > matmaturog 'sleeping' i-pa-damag'inform' > ipadpadamag 'inform, imperfecti ve'

It has been hypothesized that languages with partial reduplication also make use of full reduplication (Moravsik 1978: 328), making semantic and grammatical distinctions in the use of the two reduplicative types as seen in Nukuoro (Austronesian, Caroline Islands, Carroll 1965).

Reduplication: Form, function and distribution

13

14 (3.)

Carl Rubino Nukuoro Total Reduplication gohu dark vai water hano go gada smile ivi bone ahi fire

gohugohu vaivai hanohano gadagada iviivi ahiahi

getting dark watery diarrhea laugh skinny evening

Nukuoro Partial Reduplication seni sleep, sg. actor huge open, pi. goal ludu pick (trees) leisurely gai eat

sseni hhuge lludu gagai

sleep, pi. actor open, sg. goal pick trees frantically fish are biting

Languages that employ partial reduplication may do so in various ways. Reduplicated material is most often found at the beginning of a base, but occurs also in medial and final position. (4.)

Reduplicative Prefixes, Suffixes and Infixes Hunzib initial (N. Caucasian, Russia) CV(C) reduplication bat'iyab 'different' bat'bat'iyab 'very different' mugä-λ 'after' mu.muga-λ 'much later' (van den Berg 1995: 34) Choctaw (Muskogean, USA) medial CV reduplication tonoli 'to roll' tononoli 'to roll back and forth' binili 'to sit' bininili 'to rise up and sit down' (Kimball 1988: 440) Paumari (Arawakan, Brazil) final disyllabic reduplication a-odora-dora-bakhia-loamani-hi lpl-gather.up-REDUP-frequently-really-THEME 'we keep gathering them' (Chapman and Derbyshire 1991)

The phonological nature of the reduplicated material varies from language to language and construction to construction. Reduplicative morphemes are often characterized by the number of phonemes included in the copy, C, CV, CVC, V, CVCV, etc., the number of syllables to be reduplicated, or the number of repeated morae. In Ngiyambaa (Australian), the reduplicant consists of a copy of the first syllable and a copy of a light version of the

Reduplication: Form, function and distribution

15

second syllable, not including final vowel lengthening or a coda consonant (Donaldson 1980): magu-magu: 'around one', dhala-dhalarbi-ya (REDshine-PRS) 'to be pretty shiny'. The number of times a sequence is reduplicated is also a morphological factor in some languages, e.g. Mokilese (Austronesian, Micronesia) duplication vs. triplication: roar 'give a shudder' > roarroar 'be shuddering' > roarroarroar 'continue to shudder' (Harrison 1973). In Tigre (Semitic, Eritrea) internal reduplication of up to three internal syllables can be used. Each reduplication attenuates the meaning of the verb (Rose 2003: 114): (5.)

dagm-a: d9ga:gsm-a: ddga:ga:g9m-a: d3ga:ga:ga:gsm-a:

tell, relate tell stories occasionally tell stories very occasionally tell stories infrequently

In some cases, the morpheme type of the reduplicant will depend on other factors. In Mangap-Mbula (Austronesian, Umboi Island, New Guinea), the reduplicant of intensive constructions occurs as a prefix with bases that have a long penultimate vowel, otherwise it is suffixed (Spaelti 1997), e.g. baä.da > bad.baä.da 'you (sg.) be carrying', boozo > bozboozo 'very many,' vs. molo > mololo 'very long', posop > posopsop 'you (sg.) be finishing.' In Kinyarwanda (Bantu, Rwanda), intensive verbal reduplication is only present with bisyllabic stems. Monosyllabic verbs (and verbal stems consisting of a monosyllabic root and a stem extension) and polysyllabic verb stems do not reduplicate (Kimenyi 2002: 265). Reduplicative constructions can also be characterized as being simple, complex, or automatic. A simple construction is one in which the reduplicant matches the base from which it is copied without phoneme changes or additions. A complex construction involves reduplication with some different phonological material, such as a vowel or consonant change or addition, or phoneme order reversal. Mangarayi (Australian) has a pluralizing reduplicative construction in which the first consonant of the reduplicant is a copy of the onset consonant of the second syllable of the base followed by the rime of the first syllable, sometimes accompanied by the suffix -ji or -//.The newly created syllable does not correspond to any constituent in the original word: gurjag 'lily' > gurjurjagj-i 'having lots of Ii 1— lies'; gaifi 'child of maternal grandmother's brother' > gaijaqj-iji 'children of maternal grandmother's brother'; j-imgan 'knowledgeable person' > j-imgimgan 'knowledgeable people'; barfgal 'egg' > baqgaqgalji 'having lots

16

Carl Rubino

of eggs' (Merlan 1982). Some languages copy a short string of a root with extra material of varying lengths. In Yakan (Austronesian, Philippines), a morpheme consisting of the first consonant of the base, followed by the segment ew is used with some roots to express repetition or distributed action (Behrens 2002: 71): (6.)

labo' duddag saget

fall fall off mix

lewlabo' dewduddag sewsaget

keep on falling repeatedly fall off all mixed (several items)

Imitative reduplication in Indonesian involves the creation of a root-like form which generally does not exist independently and differs from the root by a vowel or consonant change (Macdonald and Soenjono 1967: 54). belat ganti umbang tjoreng tjerai erot

screen substitute float scratch sever crooked

belat-belit gonta-ganti umbang-ambing tjoreng-moreng tjerai-berai erang-erot

underhanded reciprocal drift to and fro full of scratches disperse zigzag

In Limos Kalinga (Austronesian, Philippines, a certain iterative construction is used consisting of the prefix maka-, a copy of the first syllable of the base, a light copy of the second (minus the final consonant, if any), and gemination of the first consonant at the affix boundary (Ferreirinho 1993:90). (8.)

maka-d-dawa-dawak maka-ng-ngina-ngina maka-s-saksa-saksak maka-l-ligwa-ligwat

keep keep keep keep

on on on on

performing the curing ceremony buying washing getting/standing

In Tuvan (Altaic, Siberia), diminutive 's' reduplication copies the entire base, except the initial consonant which is replaced by [s] in the reduplicant, e.g. pelek 'gift' > pelek-selek 'gift:DIMINUTIVE'. For bases that are vowel-initial, an onset [s] is added to the reduplicant, e.g. aar 'heavy' > aar-saar 'heavy :DIMINUTIVE' ; uuruk-suuruk 'simultaneously' (Harrison 2000). Tamil (Dravidian, India) displays a similar phenomenon where the initial CV of the reduplicated material is replaced by hi-, e.g. puli 'tiger' > puli kili 'tigers and other beasts'; maram 'tree' maram kiram 'trees and

Reduplication: Form, function and distribution

17

other growing things'; kaappi 'coffee' > kaappi kiippi 'coffee and other beverages' (Schiffman 1999: 172). Patterns such as these exist in a number of languages and are collectively referred to echo constructions. Malak Malak, an Australian language from Western Arnhem Land, employs a construction with certain one or two-syllable verb roots to denote a pluralizing effect. As in echo constructions, a separate consonant is employed with the reduplicated material, however, a smaller part of the base is actually copied. This reduplication involves the use of a liquid consonant [r], [r] or [1] between reduplicated vowels (Birk 1976: 95-96): (9.)

Singular Verb lam furk kulpat tikal karkwat

Plural Verb laram fururk kulparat tikalal kararkwarat

Gloss talk bury; enter load into a canoe lie down take out

Certain phonological processes may also take place that affect the form of the reduplicated constituent. Nias (Austronesian, Indonesia) disyllabic reduplication sometimes includes voicing, a-fusi 'white' a-vuzi-vuzi 'whitish' (Brown 2001). In Bissa (Niger-Congo, Burkina Faso), vowels are raised in a reduplicative prefix CiV[higi,er]- to form plural verbs (Prost 1950: 53): (10.)

Singular naso ta ba son

Plural nenaso tita biba suson

Gloss catch close do insult

Reduplication can also be discontinuous, in which a small segment is inserted between the reduplicant and base. In Alamblak (Sepik-Ramu), ba joins reduplicated constituents in an intensifying construction: hingnamarßa-ba-marßa-me-r (work-RED-0a-straight-REMOTE.PAST-3SG.MASC) 'he worked very well' (Bruce 1984: 165). In Dholuo (Nilo-Saharan, Kenya), the vowel a is inserted as a prefix to a reduplicated word base to express mitigation (Omondi 1982: 87): (11.)

tedo nyoro küöyo

cook yesterday sand

tedo atädä nyoro anyorä küöyö aküöyä

just cooking only yesterday mere sand

18

Carl Rubino

Automatic reduplication is reduplication that is obligatory in combination with another affix, and which does not add meaning by itself to the overall construction; the affix and reduplicated matter together are monomorphemic, e.g. the Ilocano aginCV- prefix which expresses pretense > singpet 'behave' aginsi-singpet 'to pretend to behave.' In Nez Perce (Penutian, USA), the sufffix -not/-nü:t '-less' also triggers reduplication. (Aoki 1963:43): (12.)

tohon samx

leggings shirt

titohonot sismäxnot

without leggings without shirt

Reduplicative constructions are most likely to be continuous. Reduplicative prefixes occur next to material that is copied from the initial part of a base; suffixes follow material that is copied from the end of a base. However, in Chukchi (Chukotko-Kamchatkan, Russia), some absolutive nouns are formed with a reduplicative suffix consisting of material from the beginning of the base (Dunn 1999: 108): (13.)

Singular, Abs. irw-2-ir jokwa-jow kdmPs-ksm weni-wen ίαηη-9-ίαη jilPe-jil

Plural, Abs. irw-9-t jokwa-t kdm?-5-t weni-t ίαηη-2-t jilPe-t

Gloss edged weapon eider duck worm, caterpillar bell stranger Arctic ground squirrel

Some languages may employ more than one type of reduplicative affix in the same word. Ilocano employs a construction used with onomatopoetic roots that consists of a copy of the initial consonant, followed by a replacive vowel a with another copy of the initial consonant, and accompanied by reduplication of the final vowel: CiaC]- -V fma |-, e.g. kitol /kitul/ 'sound of shoes' > kaktüol /kaktuul/ 'repeated clicking of heels'; bitog /bitug/ 'thumping sound' > babtiiog /babtuug/ 'knock down; punch'; kireb 'slamming sound' > kakreeb 'slam resoundingly' (Rubino 2001b). Finally, in some languages there are restrictions on what can appear in the reduplicant, stemming from phonological constraints or historical factors. In Tagalog, complex onset syllables occur frequently in the language from foreign loans, e.g. trabaho 'work', prutas 'fruit'. However, monophonemic onsets are preferred in reduplicants, e.g. magtatrabaho 'will work', magpuprutas 'fruit vendor.' In Malagasy, the word endings -ka, -tra, -na do

Reduplication: Form, function and distribution

19

not participate in reduplication, e.g. pitsoka 'foolish' > mi-pitso-pitsoka 'a little bit stupid, foolish', mi-petraka 'to sit' > mi-petrapetraka 'to sit about'. The deviant behavior of the stem formatives -ka, -tra, and -na reflects the fact that they are a product of a default vowel /a/ added after historically consonant-final words as part of the development towards the modern CV structure in Malagasy, e.g. volana 'moon' (< *bulan), sdratra 'writing' ( datu'datu' 'doll'; Indonesian mata 'eye' > matamata 'spy'; bantal 'pillow' > bantalbantal 'railway tie'; Tok Pisin wil 'wheel' > wilwil 'bicycle'; Ilocano tao 'human' > taotao 'pupil of the eye'; tukak 'frog' tukaktukak 'wart'; Inseno (Chumashan, USA) axmuyun 'burn, smart with pain' > axmuyuxmuyun 'have courage' (Applegate 1976: 272); Mapun (Austronesian, Philippines) sapi' 'cow' > sapi'-sapi"cowrie shell', bangkay 'corpse' > bangkaybangkayan 'period of time when relatives look over a body in state' (Collins, Collins, and Hashim 2001). With verbs (and adjectives), reduplication may be used to denote a number of things such as number (plurality, distribution, collectivity), distribution of an argument; tense; aspect (continued or repeated occurrence; completion; inchoativity), attenuation, intensity, transitivity (valence, object defocusing), conditionality, reciprocity, pretense, etc. For example, Alabama (Muskogean, USA) marks the temporary versus permanent distinction in verbal aspect with vowel lengthening loca 'to be black (covered in soot)' vs. looca 'to be a black person' as well as attenuation with gemination kasatka 'cold' > kässatka 'cool' lamatki 'straight' Icimmatki 'pretty straight' (Hardy and Montier 1988). Reduplicative inflection can be seen in many iterative or plural formations as in Hitchiti (Muskogean) adjectival stems: cikti 'thick (liquids)' >

20

Carl Rubino

cikci.ti 'thick, plural' (Kimball 1988: 440). Luiseno (Uto-Aztecan) employs two types of reduplication quite iconically to denote various plural actions: lawi 'to make a hole', law-lawi 'to make two holes, make a hole twice', lawa-lawi 'to make many holes, more than two' (Kroeber and Grace 1960), as Lampung (Austronesian, Indonesia) uses different reduplicative constructions to signal varying degrees of intensity: balak-balak 'very large', xa-xabay 'somewhat afraid' (Walker 1976). Arapesh (Torricelli, Papua New Guinea) employs reduplication to intensify or distribute the meaning of an action, often implying carelessness or lack of control on the part of the agent: su 'touch, hold' susu 'touch all over, paw'; ripok 'cut' rirtpok 'hack up' (Dobrin 2001: 36). Comox (Salish, Canada) employs -VC reduplication to express actions which lack control: c'ek'w-n Ί put a light on it' > c'ik'w-k'w-n Ί put a light on it by mistake' (Kroeber 1988: 162). Luiseno employs initial reduplication to express an emphatic conditional emphasizing that the verbal stem action may be carried out with the volition of the actor (Kroeber and Grace 1960): (14.)

nec-neci woko-woko?ax ya:-ya-yax sa:-sa-sa:msa

would would would would

certainly pay arrive go all over to tell it to all buy

Reduplication is used in a few languages to mark the inchoative, designating the start of a verbal action. Inceptive verbs in Till (Salish) are marked with double initial or final reduplication (Reichard 1959: 244): (15.)

£7 da s-li-i-fe'i-i asnux ns-i-i-as-asmxw-i yahs c-yi-yi-yahis-ύί

sick I am beginning to sicken. know I begin to know. see I begin to see.

In Alabama, actions that are imperfective in aspect (incomplete or lacking closure) appear in the language with medial reduplication, e.g. potooli 'touch' > pottooli 'coming together' (Hardy and Montier 1988: 413). Reduplication can be used to create lexical subclasses. Ilocano employs partial reduplication to form comparative adjectives, e.g. dakkel 'big' > dak-dakkel 'bigger'; na-ijisit 'dark' > naijisyisit 'darker', na-?imas 'delicious' > na-?im?imas 'more delicious.' Fijian employs full reduplication to

Reduplication: Form, function and distribution

21

derive an intransitive verb from a primarily transitive one, e.g. cula 'sew' > cula-cula 'sew away'; rabe 'kick' > rabe-rabe 'do a lot of kicking' (Dixon 1988: 48). With nouns, reduplicative morphemes have been known to denote concepts such as number, case (#13), distributivity, indefiniteness, reciprocity, size (diminutives or augmentatives), and associative qualities. For instance, Papago (Uto-Aztecan, USA) plurals: gogs 'dog' > gogogs 'dogs' (Zepeda 1983); Ilocano reciprocals (Austronesian): balem-bales (CVCN-revenge) 'avenge each other' (Rubino 2000); Nez Perce diminutives: xomayac 'mischievous child' > xoyamacxomayac 'small mischievous child' (Aoki 1963: 43); and Yokuts (Penutian) associatives k'ohis 'buttocks' > k'ok'ohis 'one with large buttocks' (Newman 1944). Reduplication is also a common method of forming indefinite pronouns, e.g. Tausug hisiyu-siyu 'whoever, anybody' from hisiyu 'who', Mapun mmooy-mmooy 'whichever' from mmooy 'which' (Collins, Collins, and Hashim 2001). With numbers, reduplication has been found to express various categories including collectives, distributives, multiplicatives, and limitatives. For example, Santali (Austro-Asiatic, India) ge-gel '10 each, by tens', Pangasinan limitatives tal-talora 'only three'; Ao Naga (Tibeto-Burman, India) final CVC reduplication distributives asem 'three' > asemsem 'three each', tinet 'seven' > tinetnet 'seven each' (Gowda 1975: 39); Javanese sanga 'nine' > sanga-sangane 'all nine' (Steinhauer 2001: 352). Reduplication is also used derivationally to alter word class, e.g. Kayardild (Pama-Nyungan) kandu 'blood' > kandukandu 'red' (Evans 1995); Luiseno (Uto-Aztecan, USA) lepi 'to tan, soften' > lepe-lpi-s 'pliable' (Kroeber and Grace 1960); Tigak (Austronesian) giak 'send' > gigiak 'messenger' (Beaumont 1979); Nama (Khoisan) causatives lom 'difficult' > !0m!om 'make something difficult' (note that the tone of the second syllable is lowered to mid tone). (Hagman 1977: 18). Full reduplication of temporal nouns is used in several languages to derive temporal adverbials, e.g. Indonesian pagi-pagi 'early in the morning' f r o m p a g i 'morning'; Tausug du:m-du:m 'every night' from du:m 'night.' Indonesian employs full reduplication of certain verbs to derive adverbials (Macdonald and Soenjono 1967: 58): (16.)

diam tiba kira masak coba

be silent arrive guess mature try

diam-diam tiba-tiba kira-kira masak-masak coba-coba

secretly suddenly at a guess maturely tentatively

22

Carl Rubino

Nez Perce employs reduplication with certain nouns to lexicalize colors, e.g. mäqs 'gall' > maqsmdqs 'yellow'; simux 'charcoal' > cimüxcimux 'black'; M:s 'water' > ku.skü.s 'blue gray' (Aoki 1963: 43-44).

3. Distribution Languages on the accompanying map are classified as having a productive reduplicative morpheme, only if the morpheme can be systematically generalized to a set of open class words, and/or the morpheme can still be applied in the modern form of the language. For example, Greek is classified as a language that does not meaningfully employ reduplication, although there are a few reduplicative forms present in the modern language that are remnants of a previously productive reduplicative process. In Ancient Greek, the perfect was formed by a Ce- reduplicative prefix, e.g. gi-grapha 'have written'; the modern equivalent is now periphrastic 'exo grapsi (have + participial form).' The old construction still appears, however, in some learned words, e.g. δβ-δό-mena (Ce-give-MEDioPASSiVE) 'data', yi-yon-os (Ce-become/happen-PERFECT) 'event.' Greek has also borrowed from Turkish a nonproductive reduplicative prefix used with at least one affective/intensive adjective: tsir-tsiplakis 'buck naked' from tsiplakis 'naked' (compare Turkish bem-beyaz 'very white' from beyaz 'white'). As can be seen from the map, reduplication is a much more pervasive phenomenon than one coming from a Western-European world view might imagine. Reduplication is very common throughout Austronesia (Pacific Islands, Philippines, Indonesia, Madagascar), Australia, South Asia, and many parts of Africa, the Caucasus, and Amazonia. In the Western Hemisphere, some language families are particularly amenable to reduplication, Salishan, Pomoan, Uto-Aztecan, Algonkian, Yuman, Sahaptian, Siouan, etc, while others are not, such as Athabascan and Eskimo-Aleut. Reduplication can be found in several areas of the world that are genetically quite diverse. One such area is the Indian subcontinent where reduplicative morphemes can be found in languages spanning several families, e.g. Indo-European, Dravidian, Austro-Asiatic and Tibeto-Burman. The Horn of Africa is yet another area where reduplication plays an important role in various languages of distinct families, e.g. Nilo-Saharan and Afro-Asiatic (Omotic, Cushitic, and Semitic) families:

Reduplication: Form, function and distribution

23

(17.)

Majang reciprocals (Nilo-Saharan, Ethiopia; Unseth 1991: 244) Root 3p reciprocal, past tim fight ti-timiikoij they fought each other jok wound jo-jokukoq they wounded each other kon help ko-koniikoij they helped each other

(18.)

Somali plurals and intensives (Cushitic; buurdn fat buurbuurän fiicän good, fine fiicfiican macaan sweet macmacaan riix push riixriix dhaqäaq move dhaqdhaqäaq

(19.)

Amharic plurals (Semitic, Ethiopia; Leslau 2000: 41) tddlaqc big, pi. tdlaq big räggim long rägaggsm long, pi. new, pi. addis new adaddis gidär calf gidadsr calves hen dorarsrt hens doro wäyzazdr ladies wäyzäro lady

Saeed 1999: 48-49) fat, pi. good, fine, pi. sweet, pi. push around move back and forth

Western Europe is one area where reduplication does not play a role in the morphology. However, Creoles that have developed from Western European languages are often found to employ reduplication quite productively, in many cases due to substratum influence, e.g. Nigerian Pidgin English kop 'cup' > kopkop 'by the cup', tüde 'today' > tude-tude 'this very day', möto 'car' > möto-mdto 'many cars', dem 'them' > demdem 'themselves (reciprocal)', tyar 'torn' > tyar-tyar 'shredded up', wäka 'walk' > wäka-wäka 'walking', trowe 'overflow' > trowe-trowe 'overflow profusely' mek 'make' > mekmek 'scheme, plot', atöl 'at all' ätol-ätöl 'under no circumstances', gbüdüm 'heavily' > gbüdum-gbüdüm 'very heavily' (Faraclas 1996: 253); Seselwa (Seychelles Creole French) ver 'green' > e rob ver-ver 'a greenish dress' > e ver-ver rob 'a deep green dress', roz 'ripe' > roz-roz-roz 'as ripe as can be' (Corne 1977: 31); Berbice Dutch Creole inga 'thorn' > inga-inga 'many thorns', mangi 'run' > mangi-mangi 'keep running' (Kouwenberg 1994).

24

Carl Rubino

(20.)

Jamaican Creole English taak talk likl little wan one biit beat, whip

(Bailey 1966: 16) taak-taak talk continuously likl-likl bit by bit wan-wan a few isolated ones biit-biit whip constantly

(21.)

Fa d'Ambö feyu fumözo mongo petu

(22.)

Zamboangueno Philippine Creole Spanish (Forman 1972: 121) kyere desire kyere-kyere desire intensely bird return bird-bird keep returning dmo boss tampa-amo-amo pretend to be boss sdbe know tampa-sabe-sabe pretend to know

Creole Portuguese (Gulf of Guinea; Post 1998) ugly fäfäyu very ugly pretty fumöfumözo very pretty slender mongomongo skinny black petupetu pitchblack

Although reduplicative morphemes are absent for most of the western branch of the Indo-European language family, reduplication is rather common in the Indo-Iranian languages of the east. (23.)

Sorani Kurdish (Iranian, pic curve dam time pda haste

Iraq; McCarus 1958: 82) pecpec zigzag dsmddm from time to time palapSl great haste

(24.)

Tajik (Iranian, Tajikistan; Rastorgueva 1963: noz coquetry nozunuz mayda small mayda-chuyda non bread nonpon

(25.)

Punjabi (Indo-Aryan, India; Bhatia 1993) xushii happiness xushii xushii nikkaa small nikkaa nikkaa paaNii water paaNii vaaNii kamm work kamm vamm

25-26) whim, caprice various small things food and the like happily very small water and the like work and the like

It is hoped that this study will reveal that although in some pockets of the world's languages, morphological reduplication is either non-existent, nonproductive, or confined to marginal word classes, most areas of the world

Reduplication:

Form, function and distribution

25

do have languages that employ reduplication productively for quite diverse purposes and with varying degrees of iconicity.

Notes 1.

Special thanks to Martin Haspelmath for producing the map from my database and allowing me to reprint it for this volume, to Brian Joseph for the Greek data and to Peter Bakker for the Fa d'Ambo examples. The database used to produce the map was originally created for the "World Atlas of Linguistic Structures" edited by Matthew Dryer, Martin Haspelmath, David Gil and Bernard Comrie.

References Aoki, Haruo 1963 Reduplication in Nez Perce. International Journal of American Linguistics 29: 42-44. Applegate, Richard 1976 Reduplication in Chumash. In Margaret Langdon and Shirley Silver (eds.), 271-283. The Hague: Mouton de Gruyter. Beaumont, Clive H. 1979 The Tigak language of New Ireland. Canberra: Dept. of Linguistics, Research School of Pacific Studies, Australian National University. Behrens, Dietlinde 2002 Yakan-English Dictionary. Manila: Summer Institute of Linguistics. Bailey, Beryl L. 1966 Jamaican Creole Syntax. Cambridge: Cambridge University Press. Bhatia, Tej K. 1993 Punjabi: A cognitive-descriptive grammar. New York: Routledge. Birk, D. B. W. 1976 The Malak Malak Language, Daly River (Western Arnhem Land). Canberra: Pacific Linguistics B-45. Brown, Lea in press Nias. In The Austronesian Languages of Asia and Madagascar. Nikolaus Himmelmann and Κ. Alexander Adelaar (eds.), London: Curzon Press. Bruce, Les 1984 The Alamblak Language of Papua New Guinea (East Sepik). Pacific Linguistics, C-81. Canberra: Australian National University.

26

Carl Rubino

Carroll, Vern 1965 The Structure of the Language of Nukuoro. Wellington, New Zealand: Polynesian Society. Chapman, Shirley, and Desmond C. Derbyshire 1990 Paumari. In Handbook of Amazonian Languages, Vol. 3, Desmond C. Derbyshire and Geoffrey K. Pullum (eds.), 161-352. Berlin/New York: Mouton de Gruyter. Collins, Millard Α., Virginia R. Collins, and Sulfilix A. Hashim 2001 Mapun-English Dictionary. Manila: Summer Institute of Linguistics. Corne, Chris 1977 Seychelles Creole Grammar. Tübingen: Verlag Gunter Narr. Dixon, Robert Μ. W. 1988 A Grammar of Boumaa Fijian. Chicago: University of Chicago Press. Dobrin, Lise 2001 Arapesh. In Encyclopedia of the World's Languages: Past and Present, Jane Garry and Carl Rubino (eds.), 31-38. New York/ Dublin: H. W. Wilson Press. Donaldson, Tamsin 1980 Ngiyambaa, the language of the Wangaaybuwan. Cambridge: Cambridge University Press. Dunn, Michael 1999 A grammar of Chukchi. Australian National University Ph.D. Thesis. Evans, Nicholas 1995 A grammar of Kayardild: with historical-comparative notes on Tangkic. Berlin/New York: Mouton de Gruyter. Faraclas, Nicholas G 1996 Nigerian Pidgin. London/New York: Routledge. Ferreirinho, Naomi 1993 Selected Topics in the Grammar of Limos Kalinga, The Philippines. Canberra: Pacific Linguistics B-109. Forman, Michael L 1972 A Study of Philippine Creole Spanish, Texts with Grammatical Analysis. Ph.D. Dissertation, Cornell University. Gowda, K. S. Gurubasave 1975 Ao Grammar. Mysore. Central Institute of Indian Languages Grammar Series I. Hagman, Roy S. 1977 Nama Hottentot Grammar. Bloomington: Indiana University Research Center for Language and Semiotic Studies. Hardy, Heather and Timothy Montier 1988 Imperfective gemination in Alabama. International Journal of American Linguistics 54: 399^115.

Reduplication:

Form, function and distribution

27

Harrison, David 2000 Topics in the phonology and morphology of Tuvan. Ph.D. Dissertation, Yale University. Harrison, Sheldon P. 1973 Reduplication in Micronesian Languages. Oceanic Linguistics 12: 407-454. Hassan, Irene U., Nurhadan Halud, Seymour A. Ashley, and Mary L. Ashley 1994 Tausug-English Dictionary: Kabtangan Iban Maana. Manila: Summer Institute of Linguistics. Kimball, Geoffrey 1988 Koasati Reduplication. In In Honor of Mary Haas, William Shipley (ed.), 4 3 2 ^ 4 2 . Berlin/New York: Mouton de Gruyter. Kimenyi, Alexandre 2002 A Tonal Grammar of Kinyarwanda — Au Autosegmental and Metrical Analysis. Lewiston: The Edwin Mellen Press. Kouwenberg, Silvia 1994 A Grammar of Berbice Dutch Creole. Berlin/New York: Walter De Gruyter. Kroeber, Alfred L. and George W. Grace 1960 The Sparkman Grammar of Luiseho. Berkeley: University of California Press. Kroeber, Paul 1988 Inceptive reduplication in Comox and Interior Salish. International Journal of American Linguistics 54: 141-67. Leslau, Wolf 2000 Introductory Grammar of Amharic. Wiesbaden: Harrassowitz Verlag. Macdonald, R. Ross and Soenjono Darjowidjodjo 1967 Indonesian Reference Grammar. Washington DC: Georgetown University Press. McCarus, Ernest N. 1958 A Kurdish Grammar: Descriptive Analysis of the Kurdish of Sulaimaniya, Iraq. New York: American Council of Learned Societies, Program in Oriental Languages, Publication B-10. Merlan, Francesca 1982 Mangarayi. Lingua Descriptive Studies 4. Amsterdam: NorthHolland. Moravcsik, Edith A. 1978 Reduplicative Constructions. In Universals of Human Languages, Joseph Greenberg (ed.), 297-334. Stanford: Stanford University Press.

28

Carl Rubino

Newman, Stanley 1944 The Yokuts Language of California. New York: Johnson Reprint Co. Omondi, Lucia Ndong'a 1982 The Major Syntactic Structures of Dholuo. Berlin: Dietrich Reimer Verlag. Post, Marike 1998 Reduplicatie in het Fa d'Ambö. In Mengelwerk voor Muysken, Adrienne Bruyn and Jacques Arends (eds.), 94-97. Amsterdam: Institut voor Algemene Taalwetenschap, Universiteit van Amsterdam. Prost, Andre 1950 La Langue Bisa: Grammaire et Dictionnaire. Ouagadougou: Centre ifan. Rasoloson, Janie N. and Carl Rubino in press Malagasy. In The Austronesian Languages of Asia and Madagascar. Nikolaus Himmelmann and Κ. Alexander Adelaar (eds.) London: Curzon Press. Rastorgueva, V. S. 1963 A Short Sketch of Tajik Grammar [Translated and edited by Herbert H. Paper], The Hague: Mouton. Reichard, Gladys A. 1959 A comparison of five Salish languages: V, International Journal of American Linguistics 25 (4): 239-253. Rose, Sharon 2003 Triple Take: Tigre and the case of internal reduplication. San Diego Linguistic Papers, Issue 1, Paper 5. Rubino, Carl 2000 Ilocano Dictionary and Grammar. Honolulu: University of Hawaii Press. 2001a Pangasinan. In Encyclopedia of the World's Languages: Past and Present, Jane Garry and Carl Rubino (eds.), 539-542. New York/ Dublin: H. W. Wilson Press. 2001b Iconic Morphology and Word Formation in Ilocano. In Ideophones, F. K. Erhard Voeltz and Christa Kilian-Hatz (eds.), 303-320. Amsterdam: John Benjamins. Saeed, John 1999 Somali. Amsterdam: John Benjamins. Schiffman, Harold F. 1999 A Reference Grammar of Spoken Tamil. Cambridge: Cambridge University Press. Spaelti, Philip 1997 Dimensions of variation in multi-pattern reduplication. University of California, Santa Cruz, Ph.D. Dissertation.

Reduplication: Form, function and distribution

29

Steinhauer, Hein 2001 Javanese. In Encyclopedia of the World's Languages: Past and Present, Jane Garry and Carl Rubino (eds.), 350-355. New York/ Dublin: H. W. Wilson Press. Unseth, Pete 1991 Reduplication in Majang. In Proceedings of the Third Nilo-Saharan Linguistics Colloquium, Franz Rottland and Lucia N. Omondi (eds.) Hamburg: Helmut Buske Verlag. Van den Berg, Helma 1995 A Grammar of Hunzib. Lincom Studies in Caucasian Linguistics 1. Munich: Lincom Europa. Walker, Dale F. 1976 Grammar of the Lampung Language. Jakarta: Badan Penyelenggara Seri NUSA [Linguistic Studies in Indonesian and Languages of Indonesia], Zepeda, Ofelia 1983 A Papago Grammar. Tucson: University of Arizona Press.

From repetition to reduplication in Riau Indonesian David Gil

1. Introduction Repetition and reduplication are superficially similar phenomena characterized by the iteration of linguistic material. By definition, repetition and reduplication differ in the following way: whereas repetition applies across words, and is therefore subsumed under syntax or discourse, reduplication applies within words, and is consequently taken to be part of morphology. Accordingly, the distinction between repetition and reduplication rests crucially on the notion of word. When the identification of words and word boundaries is clear and straightforward, the distinction between repetition and reduplication is correspondingly clear and unambivalent. However, in those cases when the identification of words and word boundaries is problematical, the distinction between repetition and reduplication may also be fraught with difficulties. One type of language in which the notion of word is problematical is the isolating type, characteristic of Southeast Asia and West Africa, though also found in other parts of the world. In languages with a substantial amount of morphology it is generally possible to identify a set of criterial features, some universal, others language-specific, distinguishing word-external syntactic structure resulting from the concatenation of words and phrases, from word-internal morphological structure resulting from processes such as affixation, compounding and the like. Such criteria form the basis for the existence, within linguistic theory, of autonomous disciplines of syntax and morphology; this point is forcefully argued by Anderson (1982). However, in isolating languages, characterized by a paucity of morphological structure, there may not be enough morphology to support a robust and systematic distinction between morphological and syntactic structure. Accordingly, in isolating languages, there may be relatively little evidence for the existence of words as a viable unit of linguistic structure, as distinct from morphemes.

32

David Gil

Thus, in isolating languages, one would expect to find cases where the distinction between repetition and reduplication is hard or even impossible to make. Within various theoretical approaches, reduplication is characterized as involving processes such as affixation, as in Marantz (1982), or alternatively compounding, as in Inkelas (this volume). But in an isolating language without clear and well-defined grammatical classes of affixes or compounds, there may be no consistent yardstick against which would-be instances of reduplication may be systematically evaluated, and distinguished from other potential instances of repetition. This paper presents a case study of the distinction between repetition and reduplication in one typical exemplar of an isolating language, namely the Riau dialect of Indonesian. After proposing a set of diagnostic criteria for distinguishing between repetition and reduplication, these criteria are applied to a wide range of data, beginning with clear cases of repetition, moving on through a variety of more problematical cases, before ending up with clear cases of reduplication. The existence of at least some clear cut cases of reduplication in Riau Indonesian suggests that, in spite of its overall isolating character, it is endowed with at least some amount of internal morphological structure.

2. Diagnostic criteria The data discussed in this paper are all spontaneous utterances from naturalistic corpora. When examining such data, it is not always clear whether a given form, exhibiting iteration, is more appropriately characterized as involving repetition or reduplication. In order to approach the problem in a systematic fashion, it is necessary to make use of a set of objective criteria distinguishing between the two. Such a set of criteria is proposed in Table 1 below:

From repetition to reduplication in Riau Indonesian

33

Table 1. Criteria for distinguishing repetition and reduplication criterion

repetition

reduplication

1

unit of output

greater than word

equal to or smaller than word

2

communicative reinforce- present or absent ment

absent

3

interpretation

iconic or absent

arbitrary or iconic

4

intonational domain of output

within one or more intonation groups

within one intonation group

5

contiguity of copies

contiguous or disjoint

contiguous

6

number of copies

two or more

usually two

The first criterion is essentially definitional. As noted in Section 1, repetition applies across words while reduplication applies within words. However, in order to make this distinction operational, one could, in principle, examine either the unit of input, that is to say, the individual pieces that are copied, or the unit of output, the structure comprising the multiple copies. In general, repetition takes as input a unit greater than or equal to a single word and yields as output a unit greater than a single word. In contrast, reduplication involves two other possibilities: total reduplication takes a single word as input and yields a single word as output, while partial reduplication takes a sequence of sounds smaller then the word as input and yields another sequence of sounds smaller than the word as output. Thus, in order to distinguish repetition from reduplication, it suffices to examine the unit of output: if it is greater than a word, we have repetition, whereas if it is equal to or smaller than a word, we are dealing with an instance of reduplication. The problem, though, with this criterion is that it presupposes the ability to identify words and word boundaries, and it is precisely such identification which is often at issue in an isolating language such as Riau Indonesian. Thus, while in many cases this criterion yields clear and unambiguous results, in many other cases it fails to provide a clear characterization of instances of iteration as involving either repetition or reduplication.

34

David Gil

The next four criteria make reference to properties which are more readily observable and which therefore may in some cases provide for a more clear cut distinction between repetition and reduplication. The first of these criteria, communicative reinforcement, pertains to a common function of repetition: speakers often repeat themselves in order to make sure that their message has been transmitted successfully. Communicative reinforcement may be necessary to overcome background noise, to achieve turn-taking in a conversation, to ensure the hearers' attention, or for many other reasons. Some examples of repetition motivated by communicative reinforcement considered below include the cries of vendors hawking their wares, and the vocative calls of speakers trying to attract other people's attention. Nevertheless, as suggested in Table 1 above, whereas communicative reinforcement is a common feature of repetition, it is not the case that all instances of repetition are motivated in this way; many other instances of repetition do not involve communicative reinforcement. As a result, when communicative reinforcement is present we may conclude that we have an instance of repetition; however, when communicative reinforcement is absent this criterion does not distinguish between repetition and reduplication. The next criterion is of a semantic nature, pertaining to the meanings, if any, associated with repetition and reduplication. Repetition is often devoid of any meaning whatsoever. (By "meaning" I mean the term in the narrow sense associated with linguistic semantics, not in any of the broader senses associated with poetics, semiotics, and other disciplines). When meaning is absent, repetition may have the function of communicative reinforcement, or it may have other functions, such as coherence (Tannen 1987) or decoration (Haiman to appear). However, in other instances, repetition is associated with meaning; in general, in such cases, the meaning in question is iconic, involving concepts such as intensivity, plurality and iterativity (König 1971, Lindström 1999, and also Moravcsik 19781). In contrast, reduplication is mostly or always associated with particular meanings; in this respect, it resembles other grammatical morphemes. However, as noted by Moravcsik (1978), Niepokuj (1991), Regier (1998) and others, the meanings associated with reduplication across languages are typically drawn from a fixed and rather limited repertoire. In this respect, reduplication differs from most other substantive grammatical morphemes, whose meanings may vary unpredictably from language to language. (For example, an -s suffix could mark nominal plurality or present third-person singular in English, direct-object first-person singular in Kashmiri, dative case in Georgian, causative voice in Iraqw, obviative in Kutenai, past tense

From repetition to reduplication in Riau Indonesian

35

in Kunuz Nubian, future tense in Tetelcingo Nahuatl, and so forth.) In particular, many of the meanings characteristic of reduplication, such as plurality, are also of an iconic nature, overlapping with those characteristic of repetition. However, many other meanings typically associated with reduplication, for example depreciation, are less obviously iconic; such meanings may accordingly be characterized as arbitrary. In addition to the cross-linguistically recurring meanings, reduplication may be associated, in particular languages, with additional, arbitrary and idiosyncratic meanings, just like most other inflectional morphemes. In Riau Indonesian, reduplication is associated with a wide range of interrelated meanings, most or all drawn from the set of meanings proposed by Moravcsik. Some of these meanings are of a high degree of iconicity; these include plurality, large quantity, large number, large size, intensivity, universal quantification, distributivity, indefiniteness, iterativity, durativity and reciprocity. However, other meanings seem to lack obvious iconic motivation; among these are atelicity, play, imitation, depreciation, concessivity, restriction and negative polarity. In addition to the above meanings, which are of a general and productive nature, there is also a small residue of more specific meanings generally associated with restricted word classes or individual words. In summary, since iconicity is characteristic both of repetition and of reduplication, in those cases where an iconic meaning is present, the semantic criterion fails to distinguish between repetition and reduplication. The cases where this criterion does differentiate between the two are the following: first, when no meaning is present, we have an instance of repetition; secondly, when a non-iconic or arbitrary meaning is present, we are faced with an example of reduplication. The next criterion pertains to a formal property of the construction in question, its interaction with the intonational structure of the utterance, and in particular, the basic unit known as an intonation group. (Alternative terms referring to a similar or identical unit include intonation phrase, melodic unit, and tone group.) In general, except for the kind of abrupt breaks caused by disfluency and similar performance factors, intonation groups span between one and a small number of whole words, which often but not always form a syntactic constituent. Intonation groups characteristically contain a single major pitch movement. Successive intonation groups are typically separated by pauses. In Riau Indonesian, intonation groups are relatively easily identifiable by means of their most salient feature: final prominence. Within each intonation group, the last syllable is accented,

36

David Gil

involving a combination of phonetic features including greater pitch variation, greater intensity and greater duration, with the latter of these features, duration, playing the most significant role. (For more detailed description and analysis of the intonational structure of Riau Indonesian, see Gil 2003, to appear.) Since intonation groups span sequences of whole words, they provide a useful criterion for discriminating between repetition and reduplication. Specifically, whereas repetition, as a multi-word construction, may sometimes straddle two or more intonation groups, reduplication, as a word-internal phenomenon, must fall within the bounds of a single intonation group. Equivalently, whereas repeated items may or may not be separated by a pause, reduplicated items are never separated by a pause. The following criterion also concerns a formal property of the construction, specifically, contiguity, or whether the iterated elements are adjacent to each other. In general, whereas repetition may involve identical copies of material that are either adjacent to each other or disjoint, separated by a stretch of other, non-identical material, reduplication, with few exceptions, involves copies that are adjacent to each other. The sixth and final criterion also pertains to a readily observable formal property, though one whose implications are not completely categorical but, rather, a matter of degree: the number of copies. In cases of repetition, there is no clear upper limit with respect to the number of copies that are present: there can, in principle, be two, three, four, or any larger number. In contrast, in cases of reduplication, the number of copies is in an overwhelming majority of cases two; this is even reflected in the term reduplication. However, although a prototypical property of reduplication, it is not exceptionless; the linguistic literature contains a smattering of cases of reduplication involving more than two copies, in languages such as Gä (Gil 1982: 321), Tigre (Rose to appear), Thao (Blust 2001), Hawaiian (Elbert and Pukui 1979: 66) and others. Such cases may be referred to as multiple reduplication, in contrast to the usual case, involving exactly two copies, of simple reduplication. Accordingly, if a particular construction involves more than two copies, we may conclude that it is likely to be an instance of repetition; however, we cannot rule out the possibility that it may involve reduplication. Indeed, as we shall see below, Riau Indonesian offers a clear cut case of a language with productive multiple reduplication. The six criteria presented in Table 1 and discussed in the preceding paragraphs are recapitulated in the form of explicit diagnostic tests in Table 2 below:

From repetition to reduplication in Riau Indonesian

37

Table 2. Diagnostics for repetition and reduplication criterion

repetition

reduplication

1

unit of output

greater than word

smaller than or equal to word

2

communicative reinforcement

present

3

interpretation

absent

4

intonational domain of output

within more than one intonation group

5

contiguity of copies

disjoint

6

number of copies

more than two copies

arbitrary

As evident in Table 2 above, all six criteria yield explicit diagnostics for repetition, whereas only two of the six provide explicit diagnostics for reduplication. The application of the above diagnostics is illustrated with respect to prototypical instances of repetition and reduplication in examples (1) and (2) below: 2 Example (1) is a vendor cry, the kind of repetitive shout addressed at passers-by by hawkers trying to sell their wares: (1)

Balai Balai

langsung turun Balai, straight descend Balai Bang Balai langsung turun bang, FAM|elder.brother Balai straight descend FAM|elder.brother Balai langsung turun bang bang, Balai straight descend FAM|elder.brother FAM|elder.brother bang Balai langsung turun bang FAM|elder.brother Balai straight descend FAM|elder.brother [At port, ticket agent trying to attract passengers for boat to Tanjung Balai] 'Balai boarding right away Balai, sir Balai boarding right away sir, Balai boarding right away sir sir, sir Balai boarding right away sir.'

38

David Gil

Example (1) above exhibits all the diagnostic criteria of repetition and none of the diagnostic criteria of reduplication. The above example consists of repetitions of two strings, Balai langsung turun and bang; clearly, with respect to each of these strings, the unit of output is greater than one word. As is characteristic of vendor cries, repetition functions to reinforce communication; in the present case the ticket agent keeps up a constant repetitive monologue, of which (1) is just a small part, in order to attract a continual stream of potential passengers, and to lure them away from other agents, competing with their own shrill monologues. Interpretation, however, is absent, since the presence of repetition does not add to the meaning of the individual items. Example (1) straddles several intonation groups, providing further support for its characterization as involving repetition. Moreover, the copies are not contiguous to each other, with the two relevant strings, Balai langsung turun and bang, exhibiting a complex interweaving pattern. Finally, each of the two strings occurs more than twice. Thus, example (1) offers as clear and as unambiguous an example of repetition as one could expect to find. In contrast, consider the following utterance: (2)

Main mony-monyet Vid play RED~monkey FAM|David [Asking to play a game on my laptop computer which involves a monkey] Ί want to play the monkey game, David.'

Example (2) above exhibits none of the diagnostic criteria of repetition but both of the diagnostic criteria of reduplication. In (2), illustrating partial reduplication, the unit of output is a sequence of sounds, mony-mony, that is smaller than a single word. Moreover, in the case at hand, the speaker wishes to refer to a computer game whose name he does not know, so he improvises a name, on the fly, by referring to a salient feature of the game, the presence of a monkey; in doing so, he makes use of one of the arbitrary or non-iconic meanings of reduplication, namely the concept of play. Thus, example (2) satisfies both diagnostics for reduplication. Conversely, it satisfies none of the remaining four specific diagnostics for repetition: communicative reinforcement is not present, the form mony-monyet does not occupy more than one intonation group, the copies of mony are not disjoint, and there are not more than two copies. Thus, example (2) contrasts maxi-

From repetition to reduplication in Riau Indonesian

39

mally with its predecessor, providing a prototypical instance of reduplication. Many cases of iteration in Riau Indonesian resemble examples (1) and (2), satisfying all of the diagnostic criteria for repetition or alternatively for reduplication. However, many other cases of iteration, perhaps a larger number, do not pattern in such a neat way. Some cases satisfy a proper subset of criteria for repetition, but none for reduplication, and thus may nevertheless be characterized as unambivalent instances of repetition. Similarly, some other cases satisfy a proper subset of criteria for reduplication, but none for repetition, and may thus still be characterized as clear cases of reduplication. However, yet other cases satisfy some diagnostics for repetition and some for reduplication, and thereby pose a problem for classification. And yet other cases satisfy none of the diagnostics for repetition or for reduplication, accordingly presenting a different kind of classificatory dilemma. In the following sections, we shall examine examples of all of these different kinds, following the expository path suggested by the title of this paper: from repetition to reduplication.

3.

Clear cases of repetition

In this section, we shall consider some examples of iteration, which, according to the diagnostics in Table 3, emerge as clear cut instances of repetition.

3.1.

Vendor cries

The first class of cases involve vendor cries, such as that already illustrated in (1) above. A similar example is provided in (3) below:

40 (3)

David Gil Aqua bang aqua aqua mineral.water FAM|elder.brother mineral.water mineral.water aqua roti, mineral.water bread roti ah roti aqua aqua, bread EXCL bread mineral.water mineral.water aqua aqua aqua aqua mineral.water mineral.water mineral.water mineral.water [Vendor on boat selling food and drinks] 'Mineral water sir mineral water mineral water mineral water bread, bread bread mineral water mineral water, mineral water mineral water mineral water mineral water'

Like (1), (3) also satisfies all six criteria for repetition. 3 Other examples of vendor cries, while satisfying a smaller number of the relevant diagnostics, nevertheless constitute unambivalent instances of repetition: (4)

Balai Balai Balai Balai Balai Balai Balai Balai Balai Balai [At port, ticket agent trying to attract passengers for boat to Tanjung Balai] 'Balai Balai Balai Balai Balai.'

In (4), the cry is uttered at great speed in a single intonation group, and successive copies of the repeated elements are adjacent to each other; thus, these two diagnostics do not apply. However, the remaining four diagnostics point clearly to the characterization of vendor cry (4) as involving repetition rather than reduplication.

3.2. Verbal art The second class of cases exemplifying repetition involves various genres of verbal art; this is illustrated with two different strategies of repetition, from prose and from poetry respectively. In Riau Indonesian, oral narratives frequently make use of a device known as tail-head linkage, in which the last few words of an utterance are repeated at the beginning of the next utterance. Whereas the first occurrence of the repeated elements is usually

From repetition to reduplication in Riau Indonesian

41

foregrounded, advancing the story line of the narrative, the second occurrence of the repeated elements is typically backgrounded, providing the setting for the continuation of the utterance, which advances the narrative one step further. Example (5) below provides an illustration of tail-head linkage from a folk tale. (5)

Sudah itu, pas satu hari, dia main judi PFCT DEM-DEM:DIST leave one day 3 play gamble Main judi 'kan, "Mat, bapak kau sakit" play gamble Q FAM|Muhamat father 2 hurt [From tale about mother and ungrateful son] 'Then after one day he went gambling. As he was gambling somebody said to him "Mat, your father's sick".'

In the above example, three of the diagnostics point unambiguously towards repetition. The unit of output is multi-word; repetition adds nothing to the meaning of the construction, and repetition straddles the intonational boundary between the first utterance and the second. The remaining three diagnostics are inapplicable. Communicative reinforcement is absent; the copies are adjacent to each other; and the number of copies is two. Moving from prose to poetry, one of the most common verse types in Malay/Indonesian is the quatrain form known as the pantun. The pantun consists of four lines, each of which in turn contains four major lexical items. In a good pantun, complex patterns of parallelism will connect the first and second couplets, making use of a variety of phonological, morphological, syntactic and semantic features; often, the first couplet will contain a seemingly irrelevant statement, setting the stage for the "message" encapsulated in the second couplet. (For a more detailed analysis of the prosodic structure of the pantun, see Gil 1993.) In many cases, the structure of the pantun is reflected by patterns of repetition. Example (6) below is quite typical in this respect, exhibiting a complex pattern involving four iterated elements, jangan suka, lebih baik, menanam and mencium. Denoting these four elements with letters A, B, C and D respectively, and other material with the symbol " ", the pattern of repetition may be represented schematically as A C ΒC AD Β D_:

42 (6)

David Gil Jangan suka NEG.IMP like Lebih more

baik good

Jangan suka NEG.IMP like Lebih more

baik good

menanam AG-plant menanam AG-plant mencium AG-kiss mencium AG-kiss

ketumbar, coriander kopi, coffee gambar, picture pipi, cheek

[Pantun] 'Don't plant coriander; Better to plant coffee; Don't kiss a picture; Better to kiss a cheek.' In the above example, four of the diagnostics point unambiguously towards repetition: the unit of output is multi-word; repetition adds nothing to the meaning of the construction; it straddles intonational boundaries; which is hardly surprising, given that it is disjoint, occurring in separate lines. The remaining two diagnostics are inapplicable: communicative reinforcement is absent; and the number of copies is two. Thus, as suggested by examples (5) and (6), various genres of stylized language provide a rich source for different kinds of repetition.

3.3. Ordinary language: The ABA construction To this point, the examples of repetition have come from various specialized language registers. But repetition is rife throughout language, including in more commonplace or everyday registers. One specific construction involving repetition is the ABA construction, in which the first word of an utterance, often a vocative expression, recurs at the end of the utterance; this construction occurs with great frequency in Riau Indonesian as in many other languages (Gil 1992). Following is an example of the ABA construction: (7)

Vid, hilangkan ini lupa dah Vid FAM|David disappear-EP DEM-DEM:PROX forget PFCT FAM|David [Playing billiards on laptop computer; speaker asking me to help him get rid of the lines on the screen which show where the balls will go] 'David, I've forgotten how to get rid of these, David.'

From repetition to reduplication in Riau Indonesian

43

In the above example, five of the diagnostics clearly characterize the construction as involving repetition. The unit of output is multi-word; the construction functions to reinforce communication, in the above case by maintaining the attention of the addressee through repetition of his name; such repetition does not contribute to the construction's meaning; intonational breaks typically occur before the Β term, as in the above example, or after it, or in both positions, as a result of which the construction involves two or more intonation groups; and the two A terms are disjoint, separated by the intervening Β term. However, the remaining diagnostic is neutral, since the number of copies is two.

3.4. Ordinary language: Reinforcing repetition Repetition in ordinary language is much more widespread than the ABA construction. Following are some typical examples of repetition associated with the function of communicative reinforcement: (8)

Ini

pencet

DEM-DEM:PROX p r e s s

ha,

ini

pencet

DEIC

DEM-DEM: PROX

press

[Playing billiards on laptop computer; showing friend which key to press] 'Press this one, press this one.' (9)

Sini

Ad,

LOC-DEM-DEM:PROX

FAM|Fuad

sini

Ad,

LOC-DEM-DEM:PROX

FAM|Fuad

sini

Ad

LOC-DEM-DEM: PROX

FAM|Fuad

[Playing billiards on laptop computer; telling friend where to shoot] 'Here Fuad, here Fuad, here Fuad.' (10)

Masuk, masuk, masuk, masuk, ah, kelok bola enter enter enter enter EXCL turn ball [Playing billiards on laptop computer; following the ball as it rolls] 'Go in, go in, go in, go in, damn, the ball curved.'

(11)

Besok jum'at, besok jum'at tomorrow Friday tomorrow Friday [Arguing about what day it is] 'Tomorrow's Friday, tomorrow's Friday.'

44 (12)

David Gil Siapa?

Siapa?

PERS-what

PERS-what

[In response to a knock on the door] 'Who's there? Who's there?' All of the above examples are characterized by at least the first four diagnostics as being instances of repetition. The unit of output in each case is clearly more than one word; the function of repetition is to reinforce communication, by making a command or an assertion more likely to be heard and acted upon; repetition does not contribute to the meaning of the utterance; and it straddles intonational boundaries, falling in more than one intonation group. In addition, example (8) exhibits another diagnostic feature of repetition, disjointness, the two copies being separated by the deictic particle ha, while examples (9) and (10) exhibit yet another characteristic property of repetition, namely, multiple copies. Thus, examples (8) - (12) all constitute clear instances of repetition.

3.5. Ordinary language: Iconic repetition The following examples also provide clear cut instances of repetition, but of a very different kind; here, the reinforcing function is absent, and instead repetition imbues the utterance with an iconic meaning involving concepts such as iterativity and durativity: (13)

Tiap malam gin-dio gitu terus 'kan, ha, every night 3 like-DEM-DEM:DIST continue Q DEIC tiap malam, tiap malam, tiap malam, tiap malam ha every night every night every night every night DEIC [From tale about a pilgrim and a young innocent girl; the pilgrim seduces the girl] 'Every night they did it, every night, every night, every night, every night.'

(14)

"Asal mak ketuk engkau tekan, ketuk tarik, ketuk origin FAM|mother knock 2 push knock pull knock tekan, ketuk tarik ketuk tekan, ketuk tarik''' push knock pull knock push knock pull [From tale about mother teaching son the facts of life; direct quotation of mother talking to son]

From repetition to reduplication in Riau Indonesian

45

' " W h e n I knock you push, I knock you pull, I knock you push, I knock you pull, I knock you push, I knock you pull. " ' (15)

(16)

(17)

"Ya, hilang", dio nyilam, nyilam, nyilam, nyilam, nyilam, EXCL disappear 3 AG-dive AG-dive AG-dive AG-dive AG-dive nyilam, tak ada dapat AG-dive NEG exist get [From tale about a pilgrim and a young innocent girl; swimming in river, they drop something, and the girl dives to look for it] " O h no, it's gone", she said, so she dived and dived and dived and dived and dived and dived, but she couldn't find it.' Setiap hari mamaknya kerja ke hutan, cari kayu, one-every day mother-ASSOC work to forest search wood jual ke kota, cari kayu, jual ke kota sell to town search wood sell to town [From tale about mother and ungrateful son] 'Every day his mother went to work in the forest looking for wood and bringing it to sell in the town, looking for wood and bringing it to sell in the town.' Dia 3

jalanjalan, terus cari abangnya 'kan, walk walk continue search elder.brother-ASSOC Q abangnya pun cari adiknya 'kan elder.brother-ASSOC even search younger.sibling-ASSOC Q [From tale about a young boy and a sparrowhawk; the young boy has fallen off a bridge over a river and becomes separated from his two older brothers] 'He walked and walked and looked for his older brothers, and his older brothers also looked for their younger brother.'

All of the above examples are characterized by at least two diagnostics as being instances of repetition. The unit of output in each case is clearly more than one word; and repetition straddles intonational boundaries, falling in more than one intonation group. In addition, example (13) exhibits another diagnostic feature of repetition, disjointness, the first occurrence of the copied element, tiap malam, being separated from subsequent occurrences by a stretch of additional material. And examples (13) - (15) exhibit yet another characteristic property of repetition, namely, multiple copies. Thus,

46

David Gil

like the ones before them, examples (13) - (17) all constitute clear instances of repetition. To this point, we have considered constructions which, according to the diagnostic criteria, emerge as clear cut cases of repetition. We shall now consider some constructions with respect to which the diagnostic criteria paint a somewhat less clear picture.

4. Probable cases of repetition In this section we consider a construction which presents an analytical challenge but which, on balance, is probably most appropriately characterized as involving repetition rather than reduplication, partial vocative repetition·. (18)

Makan berdua Vi Vi Vid eat NON.PAT-two FAM|David FAM|David FAM|David [Speaker inviting me to share a meal with him] 'Let's eat together, David.'

As evident from the above example, this construction has a vocative function. Typically, it is formed from proper names that have undergone truncation to monosyllabic forms, the function of such truncation being to express the notion of familiarity. These monosyllabic forms occur in multiple copies, typically uttered at great speed, and with a further phonological complication: the last consonant is omitted from all but the last in the sequence of copies. Prima facie, then, the above example looks like an instance of partial reduplication, such as exemplified in (2) above and (35) - (39) in Section 6 below. However, the diagnosis of this example reveals a somewhat different picture; in fact it exhibits three of the diagnostic properties of repetition. Repetition clearly functions to reinforce communication in a way reminiscent of examples (8) - (12) in Section 3; repetition does not contribute to the meaning of the utterance; and more than two copies are present. How, then, should these observations be reconciled with what seems to be a pattern of partial reduplication? Note that unlike most other instances of partial reduplication, which most often involve the initial CV- sequence of a bi- or polysyllabic word, in the present case, the host word is itself monosyllabic. This suggests that the iteration in (18) may be characterized as complete

From repetition to reduplication in Riau Indonesian

47

iteration, with subsequent deletion of the final consonant, resulting from a fast-speech process of sandhi. In other words, such examples may be characterized as involving repetition, and thereby brought in line with other similar examples of repetition considered in the preceding section.

5.

Probable cases of reduplication

To this point, all of the examples we have considered have been of repetition. We are now ready to cross the threshold and turn our attention to instances of reduplication. But our traversing of the dividing line will be less of a long-distance leap than a series of smaller steps. Just as the preceding examples of repetition seemed to approach reduplication, the next instances of reduplication will not yet be far removed from repetition.

5.1. Vivid demonstrative reduplication We shall begin with a rather specific construction type which, although also somewhat difficult to classify, is probably most appropriately characterized as involving reduplication rather than repetition. The construction in question is limited to demonstrative expressions and may be referred to as vivid demonstrative reduplication. The paradigm of demonstrative expressions is indicated in Table 3 below: Table 3. The demonstrative paradigm proximal

distal

super-distal

simple

ni

tu

*

expanded

ini

itu

*

locative

sini

situ

sana

manner

gini

gitu

*

48

David Gil

As evident from Table 3, most of the demonstrative forms are either proximal or distal; in addition, there is a single super-distal form sana.4 Limiting our attention to the proximal and distal forms, one can easily identify the following morphemes: proximal demonstrative stem ni, distal demonstrative stem tu, general demonstrative /-, locative demonstrative s-, and manner demonstrative g-. It should be noted that the morphemes /-, s- and goccur nowhere else other than with the demonstrative stems ni and tu. Perhaps because of this, forms corresponding to those in Table 3 in Standard Malay/Indonesian are sometimes considered to be unanalyzable wholes, rather than composite polymorphemic forms. However, in Riau Indonesian, there is good reason to believe that demonstratives are endowed with internal structure. And in fact, one compelling source of evidence for such structure derives from reduplication. Each of the forms in Table 3 may be repeated or reduplicated, just like any other words in the language. One instance of this can be observed elsewhere in this paper, in the reduplicated form sini-sini in (31). However, in addition to such ordinary cases, there is another more specific construction, examples of which are provided in (19) and (20) below: (19)

Ini-ni-ni-ni-ni,

dah,

dia udah

DEM-RED~DEM:PROX PFCT 3

boleh,

PFCT

can

boleh

berga-gabung

kami lah,

can

NON.PAT-RED~join

1

main bola lah

CONTR p l a y

ball

CONTR

[Looking at pictures; speaker points to person who is about to join their football team] 'There, this one here, he can already join us, playing ball.' (20)

Yang itu-tu,

gambar

itu

REL

picture

DEM-DEM:DIST

DEM-RED~DEM:DIST

[That one, that picture] 'Asking to view a particular photograph that I had taken, trying to explain which one.' Formally, vivid demonstrative reduplication involves two or more copies of the demonstrative stem ni or tu, preceded by a single instance of the morpheme /-. (Optionally, in other examples, i- may in turn be preceded by sor g-.) Semantically, vivid demonstrative reduplication is associated with a very specific meaning: it is used in order to strengthen deictic reference and make it more accessible and more vivid. Typically, it is used in contexts

From repetition to reduplication in Riau Indonesian

49

where the speaker is either pointing to some object which the hearer may have trouble identifying, as in (19), or else trying to summon up an object from the hearer's memory, as in (20). Potentially, examples such as (19) and (20) could be analyzed either as repetition, involving an expanded form ini or itu followed by one or more simple forms ni or tu, or alternatively as reduplication of ni or tu, with the output of reduplication then preceded by /-. (Such a pattern, in which clitics apply to the output of reduplication, is common; one such example is in fact present in (19), where the partially reduplicated form ga-gabung is preceded by ber-; other examples are discussed in Section 5.2 below.) Diagnosis of these constructions points tentatively towards the latter analysis. To begin with, examples (19) and (20) exhibit none of the properties of repetition, except for, in the case of (19), multiple copies; however, as pointed out in Section 2 above, this diagnostic is much weaker, since multiple reduplication is clearly attested in Riau Indonesian. Conversely, (19) and (20) would appear, perhaps, to exhibit one of the characteristic features of reduplication, namely arbitrary interpretation. Admittedly, one may be tempted to characterize the meaning of vivid demonstrative reduplication as iconic, in that it might seem to involve one of the core iconic concepts, namely, intensivity. However, a closer look at the construction suggests otherwise. In most cases of reduplication, iconic interpretations arise in a free and unconstrained fashion. For example, an ordinary reduplicated demonstrative such as itu-itu might, in different contexts, assume any combination of the following iconic meanings: 'those', 'those [large number]', 'all of those', 'that [big thing]', 'do that again and again', and others. Similar observations hold for most other words when undergoing reduplication. In contrast, forms such as itu-tu, involving vivid demonstrative reduplication, can have none of the above kinds of iconic meanings; the only interpretation they can have is the specialized "vivid" interpretation. Thus, the "vivid" interpretation may be viewed as an arbitrary and lexically-specific interpretation associated with exactly two forms, the demonstrative stems ni and tu. To the extent that the above argument holds up, it would thus seem that the construction in question is indeed more appropriately characterized as involving reduplication rather than repetition. In fact, there would seem to be another reason to analyze examples such as (19) and (20) as involving reduplication. If, contrary to hypothesis, they were instances of repetition, then, as noted above, such repetition would involve an expanded form followed by one or more simple forms in syntactic concatenation. However, there is no obvious reason why repetition

50

David Gil

should produce exactly that pattern rather than any number of other potential ones. For example, there would be no apparent reason why (19) should consist of one instance of ini followed by four of ni, rather than, say, four instances of ni followed by one of ini, or any other of the 32 logically possible combinations of ini and ni. Crucially, though, these patterns never occur: vivid demonstrative reduplication always manifests itself in the same pattern, with one instance of an expanded form followed by one or more instances of a simple form. This pattern can be accounted for straightforwardly by characterizing it as resulting from application of i- to the output of reduplication applied to one of the simple demonstrative forms ni or tu. Thus, the available evidence would seem to point towards the characterization of the construction exemplified in (19) and (20) as involving reduplication rather than repetition. 5

5.2. Complete iconic reduplication A much larger class of constructions which are difficult to classify but which, on balance, would seem to involve reduplication rather than repetition are illustrated in examples (21) - (30) in this subsection. Common to all of these examples is that they involve iconic meanings. Consider, first, the following examples: (21)

Ini, 'kan, pulau-pulau Padang DEM-DEM:PROX Q RED~island Padang [On pier, speaker points across the strait to Padang island, which stretches for almost 180 degrees from left to right] 'That's all Padang island.'

(22)

Suara voice [On a 'Why

kau putus-putus kenapa? 2 RED~cut.off why mobile phone, with a bad connection] does your voice keep getting cut off?'

Applying the diagnostics proposed in Table 2, the above examples "score a zero"; in other words, they are maximally ambivalent, exhibiting none of the properties of repetition but also none of the properties of reduplication. Equivalently, we might say that the above examples exhibit the exact array of properties that are characteristic both of repetition and of reduplication.

From repetition to reduplication in Riau Indonesian

51

Specifically, they lack the function of communicative reinforcement; they are associated with iconic meanings, large size or quantity in (21), iterativity in (22); they fall within a single intonation group; the copies are contiguous; and the copies are two in number. Of course, the distinction between repetition and reduplication is not given apriori; in principle, it should be possible to countenance the existence of constructions that are indeterminate between the two, that fall into a grey area between morphology on the one hand and syntax/discourse on the other. Nevertheless, there would seem to be two good reasons pointing towards the characterization of the above constructions as involving reduplication rather than repetition. The first argument is of an orthographic nature. In Standard Indonesian, there are three different ways for writing reduplicated forms: separated by a space, separated by a hyphen, or with the numeral "2" following the word in question. For example, the reduplicated form of anak 'child' might be written as anak anak, anak-anak or anak2. Whereas the first mode of representation is obviously appropriate not just for reduplication but also for repetition, the latter two representations are clearly specific to reduplication. Unlike Standard Indonesian, however, Riau Indonesian is a language variety without its own writing system. As a basilectal register it is most often spoken; in those relatively infrequent cases when it is committed to writing, the orthographic conventions are based on those of Standard Indonesian, but with adaptations to accommodate the different grammatical features of Riau Indonesian, and with other kinds of variation reflecting the variable levels of literacy of the writers. Nevertheless, with respect to repetition and reduplication, the orthographic conventions for Riau Indonesian are more or less the same as for the standard language. One good source of orthographic data for Riau Indonesian is provided by SMSs (short text messages) sent by mobile phone. I have recently begun compiling a corpus of SMS messages received on my own mobile phone. Most of these messages were written by semi-literate speakers of Riau Indonesian who, until the recent advent of mobile phones, rarely had occasion to write anything. Because of their low degree of literacy, the writers are, in many cases, solving linguistic problems creatively and on the fly, by having to convert spoken language into written. Thus, such SMS messages offer a valuable window into the grammatical structures of the language. In the SMS corpus, separate words may, as expected, represent instances of repetition or reduplication. Hyphenation is rarely used, possibly due to the difficulty in accessing the hyphen symbol on the mobile phone key-

52

David Gil

board. However, the numeral "2" is often used in cases that constitute clear instances of reduplication; presence of the numeral accordingly provides a diagnostic for reduplication in the SMS corpus. So what about instances of constructions such as those in (21) and (22)? Examination of the SMS corpus shows that in many cases, constructions such as these are written with the numeral "2", thereby suggesting that the writers consider them to be instances of reduplication rather than repetition. A second argument in support of the characterization of examples (21) and (22) as involving reduplication rather than repetition involves the cooccurrence of such constructions with clitics, and mirrors the analogous argument proposed in Section 5.1 above for the characterization of examples such as (19) and (20) with vivid demonstratives as involving reduplication. The following examples illustrate the interaction of iteration such as in ( 2 1 ) - ( 2 2 ) with the patient-oriented proclitic di- in (23) and the endpoint-oriented enclitic -kan in (24): (23)

Bodoh, disimpan-simpan, tak mau dibagi-bagi stupid PAT-RED~put.away NEG want PAT-RED~share [At night market, complaining about friend who was pouring his drink, bit by bit, into a bottle, instead of sharing it with his friends] 'Stupid, he's putting it away, he won't share any of it.'

(24)

Saya masuk-masukkem semua 1 :SG RED~enter-EP all [Playing billiards, speaker brags] 'I'll pocket them all.'

Again, examples (23) and (24) above are maximally ambivalent, exhibiting none of the diagnostic properties of repetition or of reduplication. However, the position of the clitics di- and -kan provides a clue as to their structure. Specifically, these clitics may be analyzed as attaching to the periphery of the single word that constitutes the output of reduplication, simpan-simpan (or bagi-bagi) and masuk-masuk respectively. Once more, as in Section 5.1, one might entertain an alternative hypothesis, involving repetition, whereby (23) would contain a sequence of two words, disimpan and simpan (or dibagi and bagi), and (24) a sequence of two words masuk and masukkan. Again, however, there is no obvious reason why repetition should produce exactly the above patterns rather than the opposite ones, simpan followed by disimpan (or bagi followed by dibagi), and masukkan followed by ma-

From repetition to reduplication in Riau Indonesian

53

suk. But while patterns such as those in (23) and (24) occur over and over again, the opposite patterns are unattested. Thus, the interaction of iteration with clitics such as in examples (23) and (24) supports the analysis of such examples as involving the attachment of a clitic to a single reduplicated word. And by extension, it provides further support for the characterization of examples (21) and (22) preceding them as also involving reduplication rather than repetition. Thus, evidence from orthography and from cliticization join forces in characterizing the examples considered in this section as involving complete iconic reduplication. With this in mind, we may now consider another class of closely related constructions. Examples (25) and (26) resemble examples (21) - (24) in all respects but one: the unit of input to the process of reduplication would appear, prima facie, to be larger than a single word: 6 (25)

Ngapain mister keASingapore-keASingapore terus A AG-what-EP white.person RED~to Singapore continue [After I say that I'll be going to Singapore in a few days time] 'Why do you keep on going to Singapore?'

(26)

Dia pelajaran duaAjam-duaAjam 3 lesson RED~two A hour [About a school which we had just walked past] 'Their lessons are two hours each.'

Like (21) - (24) before them, examples (25) - (26) above are maximally ambivalent, exhibiting none of the diagnostic properties of repetition but also none of the diagnostic properties of reduplication. In particular, they lack the function of communicative reinforcement; they are associated with iconic meanings; they fall within a single intonation group; the copies are contiguous; and the copies are two in number. Thus, the arguments proposed above in support of the characterization of examples (21) - (24) as involving reduplication rather than repetition should in principle be applicable also to examples (25) and (26). However, there is one obvious difference: examples (25) and (26) would seem to involve more than a single word in the input, and hence, ipso facto, also in the output of the process of iteration. On this basis, then, these examples might appear to involve repetition, rather than reduplication. Nevertheless, there would seem to be good reason to characterize the examples in (25) and (26) as instances of reduplication rather than repetition.

54

David Gil

To begin with, it should be noted that the apparent phrasal nature of the above examples is an artefact of their orthographic representation. As pointed out in endnote 2 above, the examples are presented in the conventional orthography for Standard Indonesian, which is potentially misleading on two counts: first because it is an orthographic convention which may or may not correspond to the actual facts of the language, and secondly because it is meant to represent the facts of the standard language, not of Riau Indonesian. Thus, in order to evaluate the above examples, it is necessary to analyze their actual grammatical structure in Riau Indonesian. In fact, work in progress suggests that the above examples may be in fact be characterized as involving the reduplication of a single, albeit sometimes rather complex word. Example (25) may be argued to be a bimorphemic word consisting of a stem preceded by a proclitic or prefix, the directional marker ke.7 And example (26) may be analyzed as a bipartite compound word consisting of a numeral followed by a classifier or classifier-like expression, the mensural time expression jam 'hour'. Thus, in both examples, a reasonably good case could be made that the reduplicated unit consists of a single word, involving either a proclitic or a compound. Crucially, more complex strings than those typified above almost never occur as the input to reduplication. In this respect, the above examples differ crucially from several of the instances of repetition considered above, which involve the iteration of more complex expressions; see, for example, (1) and (16). Thus, the above examples may be characterized as involving the reduplication of a single polymorphemic word. In doing so, they also suggest that the constraint on how big an expression can be reduplicated might be reduced to independent constraints on word formation, via cliticization or compounding. Further support for this analysis is provided in Section 6.1 below, in the discussion of examples (33) and (34), which are shown to constitute clear cut instances of reduplication involving compound words. In conclusion to this section, we shall examine one more class of constructions, similar to those in (21) - (26), while differing only in the presence of multiple copies: (27)

Naik ojek-ojek-ojek aja go.up RED~motorcycle.taxi just [Group of three persons debating whether to take a single taxi together, or separate motorcycle taxis] 'Let's just ride motorcycle taxis.'

From repetition to reduplication in Riau Indonesian (28)

Kalau si Pai ambil-ambil-ambil-ambil TOP PERS Pai RED~take [Complaining about friend's behaviour] 'Pai just takes things all the time.'

(29)

Digalinya kuburan ya Vid, digali-gali-gali, PAT-dig-ASSOC tomb-AUG yes FAM|David PAT-RED~dig sampai lah arrive CONTR [From tale about mother and ungrateful son; after mother dies, the son digs up her grave to get hold of a ring still on her finger] 'So he dug up the grave, David, he dug and dug and dug and then he reached it.'

(30)

Dia

pencet

dia

3

press

3

55

aja just

banyak-banyak-banyakkim RED-much-EP

[About a dishonest taxi driver and his meter] 'He pressed a knob and made it show lots and lots.' Like (21) - (26), examples (27) - (30) lack the function of communicative reinforcement; they are associated with iconic meanings, — distributivity in (27), iterativity in (28), durativity in (29), and large quantity in (30); they fall within a single intonation group; and the copies are contiguous. In these respects they are indeterminate with respect to the distinction between repetition and reduplication. However, the number of copies is greater than two, which, as suggested in Section 2, points weakly towards an analysis as involving repetition. But as noted therein, examples of multiple reduplication have occasionally been attested in the literature; moreover, as shown in Section 6 below, there are clear and unambiguous instances of multiple reduplication in Riau Indonesian. Therefore, it would not be unwarranted to override this criterion should reason for doing so present itself. And in fact there may be good reasons for doing so. One reason, perhaps not a very strong one, is simply for uniformity. Examples (27) - (30) resemble examples (21) - (26), analyzed above as involving reduplication, in so many respects that it would seem rather unfortunate to have to separate them off and characterize them as involving repetition rather than reduplication solely on the basis of one relatively unimportant formal property, the number of copies. A second, perhaps stronger argument, provided by examples (29) and (30), involves the interaction of iteration and cliticization; this is the argu-

56

David Gil

ment that has already been invoked twice before, in Section 5.1 for examples (19) and (20), and earlier in this section for examples (23) and (24). In (29), the proclitic di- attaches to the beginning of the single word that constitutes the output of reduplication, gali-gali-gali; similarly, in (30), the enclitic -kan attaches to the end of the single reduplicated word banyakbanyak-banyak. If, contrary to hypothesis, these were instances of repetition, then there would be no reason why repetition should produce exactly the above patterns, with the proclitic di- at the beginning and the enclitic -kan at the end; rather, we would expect to find one or more instances of the clitics randomly interspersed amongst the repeated occurrences of the putative separate words, for example gali gali digali, gali digali digali, and so on. But this never happens; the pattern is always as in (29) and (30). Thus, the interaction of iteration with cliticization in such examples supports their characterization as involving the attachment of a clitic to a single multiply-reduplicated word. And by extension, it provides further support for the characterization of examples (27) and (28) preceding them as also involving reduplication rather than repetition. Thus, on the basis of the above two arguments, we may tentatively conclude that examples (27) - (30) provide instances of multiple reduplication. Further support for the above analysis is provided in Section 6 below, in the discussion of examples (36) and (39), which are shown to constitute unambivalent instances of multiple reduplication. Thus, examples (27) - (30) may be grouped together with examples (21) - (26) considered earlier in this section as instances of complete iconic reduplication.

6.

Clear cases of reduplication

The preceding two sections were concerned with constructions for which the diagnostic criteria fail to discriminate clearly between repetition and reduplication. In particular, the preceding section dealt, at some length, with constructions which, it was argued, in spite of the indeterminacy, should nevertheless be characterized as probable instances of reduplication. Having traversed the route from clear repetition to probable repetition to probable reduplication, we are now approaching the end point of our journey, in which we turn our attention to the many clear, unambivalent cases of reduplication.

From repetition to reduplication in Riau Indonesian

57

6.1. Complete arbitrary reduplication The first class of examples of clear reduplication resemble the examples in (21) - (26) but with one crucial difference: whereas the above examples are iconic, the examples below are associated with non-iconic meanings:

(31)

Anton

tak

sini-sini,

a

Anton

NEG

RED-LOC-DEM-DEM:PROX

EXCL

[Commenting about his friend] 'Anton hasn't been around.' (32)

Kecil-kecil punya cewek itu RED~small have girl DEM-DEM:DIST [About his little brother] 'Even though he's small, he's got girlfriends.'

(33)

TehAobeng-tehAobeng tak usah lah RED~tea A screwdriver NEG NEG. IMP CONTR [Group of people getting together an order for take-away food; speaker tries to cut down on size of order] 'No need for iced tea.'

(34)

EnamApuluh-enamApuluh biar lah Vid A RED~six ten let CONTR FAM|David [In hotel, I wonder where to do my laundry; speaker says I should have it done in the hotel; I say last time I did it there six pieces cost sixty thousand, which is too expensive; speaker responds] 'Sixty or whatever, doesn't matter, David.'

Examples (31) - (34) exhibit the crucial diagnostic feature of arbitrary interpretation; in each of these examples, reduplication is associated with an interpretation that is clearly non-iconic. In (31) and (33) reduplication is a negative polarity item, licensed by the negative marker tak and, in the latter case, also by the negative imperative marker usah, while in (32) and (34) it has a concessive denotation. On the basis of their arbitrary, non-iconic interpretations, the above examples may accordingly be characterized as clear instances of reduplication. Note, in particular, that in examples (33) and (34), the unit of input to reduplication is one that would seem to consist of two words. These exam-

58

David Gil

pies accordingly resemble those in (25) and (26), in Section 5.2, which also appeared to involve two or more words in the unit of input. However, just as in the previous examples, it would seem to be the case that what is being reduplicated here is actually a single compound word. Thus, in (33), teh obeng, literally 'tea screwdriver', is a semantically opaque compound meaning 'iced tea'. 8 And in (34), enam puluh is a complex numeral. But in (33) and (34) above, the arbitrary interpretation points clearly towards their characterization as instances of reduplication, which in turn conclusively supports their analysis as involving compounding. And this, in its own turn, provides further justification for the corresponding analysis for examples (25) and (26) in which reduplication is associated with iconic meanings.

6.2. Partial iconic reduplication Whereas examples (31) - (34) differ from (21) - (26) in that they involve arbitrary interpretations, the second class of examples of clear cut reduplication, considered below, differ from (21) - (26) in that they exhibit partial reduplication. 9 (35)

Bajunya ko-koyak clothes-ASSOC RED~tear [About my shirt] 'Your shirt's all frayed.'

itu DEM-DEM:DIST

In (35) above, the interpretation is iconic rather than arbitrary, involving spatial distributivity. However, since this example involves partial reduplication, the unit of output is actually smaller than a single word; hence, this example may be characterized as a clear instance of reduplication. The following example differs from the above one in just one respect, namely, that it involves multiple reduplication: (36)

Udah si-si-siapl PFCT RED~ready [Getting ready to go out, speaker hurries friend] 'Is everything ready?'

In (36), too, the interpretation is iconic rather than arbitrary, involving universal quantification. However, the fact that the unit of output is smaller

From repetition to reduplication in Riau Indonesian

59

than a single word forces its analysis as involving reduplication. This in turn provides conclusive evidence in support of the existence of multiple reduplication in Riau Indonesian. And this then provides further retroactive support for the characterization of examples (27) - (30) in Section 5.2 as also involving multiple reduplication.

6.3. Partial arbitrary reduplication The third class of examples of clear reduplication, considered below, combines the two diagnostics of the two preceding classes; these are examples of partial arbitrary reduplication, the most prototypical, or "best" instances of reduplication. Thus, we have finally completed the journey from prototypical repetition, as exemplified in (1), to prototypical reduplication, as exemplified in (2). Some additional examples, resembling (2), of partial arbitrary reduplication are provided below: (37)

Aku tak fa-faham lah 1:SG NEG RED~understand CONTR [Interlocutor says he's thirsty several times, hinting to speaker that he make coffee; finally, the penny drops, and the speaker says, laughing] Ί didn't understand.'

(38)

Saya ti-tidur, say a tahu 1 :SG RED~sleep 1:SG know [Describing how he slept beside his girl friend, who began to grope him] Ί was only pretending to sleep, I knew what she was doing.'

Examples (37) and (38) exhibit both of the diagnostic features of reduplication. The unit of output is smaller than a single word, and the interpretation is clearly non-iconic or arbitrary: in (37) reduplication is a negative polarity item, licensed by the negative marker tak; and in (38) it is associated with a play meaning. Thus, alongside (2), examples (37) and (38) provide the clearest instances of reduplication in Riau Indonesian. Finally, example (39) below illustrates partial arbitrary reduplication that is also multiple:

60 (39)

David Gil A,

kalau

EXCL TOP

bayar pay

orang

itu,

sikit

p e r s o n DEM-DEM:DIST a.little

cu-cu-cu-cu-cu-culit, RED~pull

mahal, kalau orang perbaiki ini expensive TOP person CAUS-good-EP DEMDEM:PROX

[About CD-player repairmen] 'Those people just fiddle a little bit and then you pay a lot, the people who fix things.' Again, in spite of the presence of multiple copies, (39) is a clear instance of reduplication. The unit of output is smaller than a single word, and the interpretation is clearly arbitrary, expressing atelicity. In fact, with its six copies of the syllable cu-, cu-cu-cu-cu-cu-culit is the instance of multiple reduplication with the greatest number of copies in the corpus, and, to the best of my knowledge, a current world record.

7. Conclusion The preceding four sections took us on a long and variegated journey from clear cases of repetition through a range of problematical cases all the way to clear cases of reduplication. In spite of the analytical problems involved, each of the problematical cases was nevertheless classified as an instance either of repetition or of reduplication. The existence of many clear cases of reduplication, in conjunction with the proposed analysis of many of the more problematical cases as also involving reduplication, together show that in spite of its generally isolating character, Riau Indonesian is indeed endowed with word-internal morphological structure, albeit perhaps of a rather impoverished nature.

Acknowledgements This paper would not have been possible without the many repetitious and reduplicating speakers of Riau Indonesian who provided the naturalistic data on which this paper is based: Aidil, Aliawar, Andi, Arief, Arip Kamil, Doni, Elly Yanto, Fuad, Ijal, Kairil, Nasar, Pai, Per, Rudy Candra, Saiful, Saripudin, Septianbudiwibowo, Suriyadi, Yonky, Zainudin, and anonymous

From repetition to reduplication

in Riau Indonesian

61

others. T h a n k s are also due to Bernard Comrie, John H a i m a n and Bernhard Hurch for helpful c o m m e n t s on previous versions of this paper, and to participants of the G r a z Reduplication Conference for s o m e u s e f u l remarks and suggestions. This paper is an abbreviated version of a m u c h longer paper titled " F r o m Repetition to Reduplication in the Riau Dialect of Indonesian", w h i c h presents a substantially larger body of data, and also examines s o m e of the diachronic implications.

Notes 1.

2.

3.

Moravcsik (1978) is often cited as the seminal study of reduplication; however, her use of the term reduplication is actually broader than that of most other scholars, subsuming also many constructions which would generally be considered as involving repetition, for example the English He is very very bright. Data from the naturalistic corpus is presented in four lines: transcription, interlinear gloss, description of the context in which the utterance occurred, and a free translation into idiomatic English. The transcription makes use of the conventional writing system of Standard Indonesian; in particular, word boundaries are in accordance with the standard orthography, even in those cases, discussed in Sections 5 and 6 below, where these diverge from the linguistic reality of Riau Indonesian. Repeated expressions are separated by spaces, while reduplicated words and smaller units are separated by hyphens. In each example, the instance or instances of repetition and reduplication under consideration are italicized; however, some examples happen to contain additional instances of repetition or reduplication, which are not italicized. Boundaries between intonation groups are represented with commas. In the interlinear glosses, repetition is represented with repeated glosses, while reduplication is represented with the symbol RED. In addition, the interlinear glosses make use of the following abbreviations: AG agent orientation; ASSOC associative; AUG augmentative; CAUS causative; CONTR contrastive; DEM demonstrative; DEIC deictic; DIST distal; EP end-point orientation; EXCL exclamation; FAM familiar; IMP imperative; LOC locative; NEG negative; PAT patient orientation; PERS personal; PFCT perfect; PROX proximal; PST past; Q question; REL relative; SG singular; TOP topic; 1 first person; 2 second person; 3 third person. Example (3) differs minimally from example (1) with respect to the unit of input to repetition; whereas in (1) it is multi-word, in (3) it is a single word. However, as noted in Section 2, this property does not figure in the diagnostic criteria specified in Table 2.

62 4.

5.

6.

7.

8.

David Gil The reasons for including sana in the paradigm and thereby introducing three additional empty cells are two-fold. First, the form sana would appear to share the locative morpheme s- argued below to be part of the forms sini and situ. Secondly, in other dialects of Indonesian, some of the remaining empty cells are filled. For example, in Jakarta Indonesian, the proximal and distal forms are identical to those in Riau Indonesian; however, in addition, there are three super-distal forms, no, ono, and sono, filling the first three cells in the final column of Table 3. Yet another argument for the characterization of the construction in question as involving reduplication derives from the comparison with the Jakarta dialect of Indonesian. In Jakarta Indonesian, reduplication differs from its counterpart in Riau Indonesian in one important respect: multiple reduplication is rare or absent; the number of copies is always, or nearly always, two. Now in Jakarta Indonesian there is a construction that resembles (19) and (20) in form and meaning; however, in this construction, the number of copies of the demonstrative stem is always, or nearly always, two. Of course, in Jakarta Indonesian, as in most or all other languages, repetition can involve any number of copies. Thus, the restriction of the construction in question in Jakarta Indonesian to two copies suggests that in Jakarta Indonesian the construction involves reduplication rather than repetition. And this, in turn, increases the plausibility of the analogous analysis for the corresponding construction in Riau Indonesian as well. In examples (25) and (26) and subsequently, a wedge symbol " Λ " is used to connect forms which in the standard orthography are written as separate words but which group together to constitute the input to the process of reduplication. As noted in the introduction, arguments for wordhood are hard to come by in an isolating language like Riau Indonesian; nevertheless, there is some evidence that ke may be a proclitic or prefix. One argument is orthographic: in the corpus of SMS messages discussed earlier, ke is often written joined on to the following word, even though the standard writing system has it written as a separate word. A second argument, of a phonological nature, is specific to one of two subdialects of Riau Indonesian, that in which there is a phonemic schwa. In that subdialect, ke contains a schwa; however, there are no other words that end with a schwa, and no other monosyllabic words whose only vowel is a schwa. Thus, for this subdialect at least, phonotactic considerations also support the analysis of ke as a proclitic or prefix. The expression teh obeng is geographically restricted to the islands of the Riau archipelago, and does not occur on the Sumatran mainland. Its origin is probably in the speech of the Chinese Singaporeans who frequently visit these islands. The Indonesian word obeng 'screwdriver' is a reinterpretation of a phonetically similar sequence in one of the Southen Min languages, whose

From repetition to reduplication in Riau Indonesian

9.

63

meaning is 'crow ice', where 'crow' is used in bisyllabic compounds to denote objects that are black, such as, in the case at hand, a drink without milk. Most commonly, partial reduplication applies to the initial CV- sequence of the host word, though occasionally it may apply to larger sequences of sounds, such as in example (2). Although less frequent than total reduplication, partial reduplication is quite productive, and in apparent free variation with total reduplication: no systematic differences in function would seem to exist between the two.

References Anderson, Stephen R. 1982 Where's morphology? Linguistic Inquiry 13: 571-612. Blust, Robert 1998 Ca-reduplication and Proto-Austronesian grammar. Oceanic Linguistics 37: 29-64. Blust, Robert 2001 Thao triplication. Oceanic Linguistics 40: 324-335. Brown, Lea 2001 A grammar of Nias Selatan, Ph.D. diss., University of Sydney, Sydney. Elbert, Samuel H. and Mary Kawena Pukui 1979 Hawaiian Grammar. Honolulu: The University Press of Hawai'i. Gil, David 1982 Distributive Numerals. Ph.D. diss., University of California, Los Angeles. Gil, David 1992 The heterogeneity of linguistic structure: Prosody, algebra and grammar in Tagalog vendor cries. In Proceedings of the 1992 Seoul International Conference on Linguistics, 361-375. Seoul: The Linguistic Society of Korea. 1993 Ί1 pleut doucement sur la ville': The rhythm of a metaphor. Poetics Today 14: 49-82. 2003 Intonation does not differentiate thematic roles in Riau Indonesian. In Proceedings of the Ninth Annual Meeting of the Austronesian Formal Linguistics Association (AFLA9), A. Riehl and T. Savella (eds.), 64-78. (Cornell Working Papers in Linguistics 19). Intonation and thematic roles in Riau Indonesian. In Topic, Focus forthc. and Intonation [tentative title], M. Gordon, D. Biiring and C.M. Lee (eds.), Deventer: Kluwer.

64

David Gil

Haiman, John forthc. Decorative imagery, unpublished paper. Inkelas, Sharon this vol. Morphological Doubling Theory: Evidence for morphological doubling in reduplication. König, Ekkehard 1971 Kumulative Komparative, Beiträge zur Generativen Grammatik, Schriften zur Linguistik 3: 100-111. Lindström, Jan 1999 Vackert, vackert! Syntaktisk reduplikation i svenskan, Ph.D. diss., University of Helsinki, Helsinki. Marantz, Alec 1982 Re reduplication, Linguistic Inquiry 13: 35^18. McGregor, William 1990 A Functional Grammar of Gooniyandi. Amsterdam: John Benjamins. Moravcsik, Edith A. 1978 Reduplicative constructions. In Universals of Human Language, Vol. 3: Word Structure, Joseph H. Greenberg (ed.), 297-334. Stanford: Stanford University Press. Niepokuj, Mary K. 1991 The historical development of reduplication, with special reference to Indo-European. Ph.D. diss., University of California, Berkeley. Regier, Terry 1998 Reduplication and the arbitrariness of the sign. In Proceedings of the Twentieth Annual Conference of the Cognitive Science Society, M. Gernbacher and S. Derry (eds.), 887-892. Mahwah, NJ: Lawrence Erlbaum Associates. Rose, Sharon forthc. Triple Take: Tigre and the case of internal reduplication. In Studies in Afroasiatic Grammar, R. Hayward, J. Ouhalla and D. Perrett (eds.), Amsterdam: John Benjamins. Tannen, Deborah 1987 Repetition in conversation: Toward a poetics of talk. Language 63: 574-605.

Morphological Doubling Theory: Evidence for morphological doubling in reduplication Sharon Inkelas

1.

Introduction*

This paper introduces and motivates Morphological Doubling Theory by focusing on the morphological evidence for viewing reduplication as a morphological construction whose daughters are constrained to be morphosemantically identical. In Morphological Doubling Theory, reduplication is viewed as the double (or multiple) occurrence of a morphological constituent meeting a particular morphosemantic description. Morphological Doubling Theory thus departs from previous theories in which the reduplicant is treated as an abstract morpheme, RED, whose substance is provided by phonological copying (e.g. Marantz 1982, Steriade 1988) or correspondence (e.g. McCarthy and Prince 1995). In Morphological Doubling Theory, reduplicant and base are both generated by the morphology as part of a construction which also embodies semantic and phonological generalizations about the output of reduplication: (1)

Mother (meaning = some function of the meaning of the daughters; phonology = some function of the phonology of the daughters)

Daughter #1 (meaning = that of Daughter #2; may be subject to special phonology)

Daughter #2 (meaning = that of Daughter #1; may be subject to special phonology)

66

Sharon Inkelas

The same morphological structure is assigned to partial and to total reduplication, which differ solely in whether or not one of the daughters is phonologically truncated.

2.

Phonological identity approaches

Previous approaches treat reduplication as phonological copying motivated by the need to segmentally flesh out a skeletal RED morpheme. The examples below, using the Chumash form for 'islanders' (Applegate 1976), illustrate preposed partial reduplication. Example (2) represents the derivational approach taken in the 1980's (e.g. Marantz 1982, McCarthy and Prince 1999, Steriade 1988); RED, a skeletal bimoraic syllable, is fleshed out by copying the base segments and associating the copies by rule to RED. Any leftover segments are stray-erased. In (3), representing the more recent Base-Reduplication Correspondence Theory (BRCT) approach to reduplication in the Optimality Theory literature (e.g. McCarthy and Prince 1993, 1995), RED is a morpheme constrained, by R e d = ^ ( ( , to instantiate a bimoraic syllable and, by FAITHBR, to correspond segmentally to the material in the base. FAITHio » FAITHBR prevents the base from truncating. (2)

Derivational approach: reduplication by copying Affixation Ο μ μ

+

Ο μ

Copy & Association

Ο μ μ

Λ

*

Ο μ μ

Α

Α

c h umas (3)

Ο μ

Ο μ μ

Stray Erasure *

Ο μ μ

Λ Α

c h umas

Λ

c h umas

Ο μ

Λ

Ο μ μ

Α

c h um-c h umas

Correspondence Theory: reduplication by correspondence RED, c h umas a. b. c.

FAITH I O

RED = σ μ μ

c h um-c h umas h

as

h

c umas- c umas h

h

c um- c um

FAITHBR

as!

MI^MLMLI

as!

For the remainder of this paper, phonological doubling theories will be represented by BRCT, as spelled out in McCarthy and Prince 1995.

Morphological

3.

Doubling Theory

67

Overall motivation for MDT

The argument for Morphological Doubling Theory (MDT), also articulated in Inkelas and Zoll (to appear), is as follows: a) b) c)

MDT is necessary: there are data that must be analyzed using MDT and which are inconsistent with BRCT. MDT is sufficient: there are no data that require BRCT for their analysis and which are inconsistent with MDT. MDT is more restrictive than BRCT, which overgenerates, predicting unattested types of reduplication patterns.

If these arguments hold up, Occam's razor supports choosing Morphological Doubling as the analysis of all cases of reduplication. The present paper focuses on the first of these three arguments. The second is too lengthy to be attempted here, though see Inkelas and Zoll (to appear). The third argument appears briefly towards the end of the paper.

4.

Comparison

The key assumption of MDT is that daughters in a reduplication construction are semantically identical. Phonological identity is not presupposed or required. As methods of generating reduplicated structures, MDT and BRCT differ in a number of important ways, summarized below: In MDT, the reduplicant is a potentially complex morphological constituent; in BRCT the reduplicant is monomorphemic In MDT, the meaning of a reduplication construction is a property of the mother node, i.e. of the construction as a whole; in BRCT, the reduplicant morpheme has its own fixed meaning In MDT, identity between base and reduplicant is semantic; in BRCT, it is phonological In MDT, the reduplicant and base are derived from phonologically independent inputs; in BRCT, a single phonological input generates both reduplicant and base

68

Sharon Inkelas In MDT, the reduplicant and base have phonologically independent outputs; in BRCT, reduplicant and base are linked on the surface by phonological correspondence

This paper will focus on the first three of these points in arguing for MDT.

5.

The morphological argument for MDT

This paper argues on the basis of morphological evidence for the need to model reduplication as morphological doubling. The argument is structured as follows. First, as documented in §6, the existence of affix reduplication shows that reduplication can target morphological subconstituents of a word, regardless of phonological size, confirming that what is doubled in reduplication is a morphosemantically defined constituent. Second, as argued in §7, the existence of synonym and antonym constructions shows that grammatical constructions must be able to require semantic similarity (or dissimilarity) of their daughters. Any grammar that can generate nonreduplicative constructions with this ability already has the ability to generate reduplicative constructions as well, identity being merely a special case of similarity. The argument that reduplication is morphological doubling is completed by the evidence, in §8, that morphological reduplication constructions exist in which base and reduplicant can be quite different phonologically, even to the point of containing different morphemes, as long as they are equivalent semantically; similar evidence from syntactic reduplication is provided in §9. The data discussed in these sections are incompatible with phonological copying approaches to reduplication.

6.

Affix reduplication

Ordinary total or partial reduplication of the root or stem of a word, as in the Chumash example mentioned earlier, is generally compatible with both morphological doubling (MDT) and phonological copying approaches to reduplication. Affix reduplication provides a better test of their differences. We examine here several cases in which what reduplicates is not the entire stem which is input to the reduplication process, nor any phonologically defined subpart of that stem, but instead a particular affix within that stem.

Morphological Doubling Theory

69

These data support the claim that reduplication is doubling of a morphological constituent, as in MDT. In Amele (Papuan; Roberts 1987, 1991: 128-29), simultaneous action in verbs is marked by CV reduplication of the verb root (4a) or of an object suffix {-do) (4b), as shown below: (4)

a. b.

bi-bil-en gba-gbatan-en abul-do-do-n mele-do-do-n

'as 'as 'as 'as

he he he he

(bil-e? sat' split' (gbatan-e? struggled' (abul-do-? examined' (mele-do-?

'to 'to 'to 'to

sit') split') struggle') examine')

In Dyirbal (Dixon 1972: 242), nominals are pluralized by reduplicating either the root (5a) or a stem-forming suffix (5b). Root reduplication and affix reduplication have the same semantic effect (5c): (5)

a. b. c.

midi-midi gulgij;i-gulgi(;i midi-bac|un-bac|un (bayi) ya|;a-gabun-gabun midi-midi-bactun (bayi) yaf_a-ya|;a-gabun

'lots 'lots 'lots 'lots 'lots 'lots

of of of of of of

little ones' prettily painted men' very small ones' other men/strangers' very small ones' other men/strangers'

If we assume that root + object marker, in Amele, and root + stem-forming suffix, in Dyirbal, form a morphological constituent that we can call a 'stem', then the generalization is the same in both languages: reduplication doubles some morphological constituent within the stem, irrespective of phonological size or linear position. These cases are not easily described in a framework that attributes reduplication to a particular morpheme with a fixed meaning. The reduplicated elements in Amele and Dyirbal don't look like RED morphemes. They don't have a common phonological shape within each language; in Amele their linear position isn't even the same. Moreover, the meaning of these reduplication constructions is a property not of the stem-internal morpheme which happens to reduplicate in any given instance, but rather of the construction as a whole. In all of these respects affix reduplication is best described as morphological doubling. Further evidence that reduplication is morphological constituent doubling, rather than arising from the use of a phonologically empty reduplicative morpheme with its own semantics, comes from triplication, the phenomenon whereby an element is repeated not just once but twice. In Mokilese

70

Sharon Inkelas

(Harrison 1973, 1976), verbs form their statives through suffixing CVX reduplication (e.g. kadip 'to betray', kadipdip 'treacherous'; Harrison 1973: 424), and their progressives through prefixing CVC reduplication (e.g. kapang 'to see', kapkapang 'watching'; Harrison 1973: 425). For monosyllabic verbs only, progressives reduplicate twice, resulting in triplication. Transcriptions are orthographic: Denotative

Progressive

gloss

kang doau soal daun jahk

kang-kang-kang doau-doau-doau soal-soal-soal dah-dah-daun jah-jah-jahk

'eat' 'climb' 'black' 'fill' 'bend'

Harrison considers the possibility that triplication is an anti-homophony strategy aimed at keeping progressive and stative forms of monosyllabic forms distinct; however, he rejects this idea on the evidence that some verbs that undergo progressive triplication are inherently stative and lack a derived stative form. Thus triplication operates even when homophony is not an issue. As Harrison concludes, triplication appears to be a requirement of the progressive construction when the input is monosyllabic. If reduplication is due to the presence of a meaningful morpheme RED, the double use of RED would be expected to correlate with a semantic change. By contrast, treating reduplication as a construction, with prespecified semantics, that simply requires one of its daughters to appear twice (or thrice) makes no such prediction. The existence of semantically vacuous triplication supports MDT.

7.

Identity effects

A number of morphological constructions require semantic identity, semantic similarity or (in some cases) semantic dissimilarity between their daughters. These constructions are not normally called reduplicative (though see Singh 1982 on Hindi), since the daughters can in some cases differ semantically as well as, crucially, phonologically. However, as we argue, any theory with the ability to model these constructions already has the ability

Morphological

Doubling Theory

71

to model reduplication, and does not need recourse to extra mechanisms like a RED morpheme or base-reduplication correspondence. Khmer (Ourn and Haiman 2000: 485, 500-502) and Vietnamese (Ourn and Haiman 2000, Nguyen 1997: 67, 70) both exhibit what we may call 'synonym compounding' in which the two members of the compound are phonologically distinct, perhaps etymologically distinct synonyms (e.g. peel-weeliQ 'time', from Sanskrit peel 'time' + Pali weelia 'time'; Ourn & Haiman, p. 485). The meaning of these constructions in Khmer can be lexicalized, as in the first two lines of (7a), but frequently are the same as the meaning of the individual parts. (7)

Khmer synonym compounds cah-tum kee-mordok camnqj-?ahaa(r) ?aar-kambaq cbah- prakat

'old + mature' 'heritage + heritage' 'food + food' 'secret + secret' 'exact + exact'

'village elder' 'legacy' 'food' 'secret' 'exact'

Vietnamese synonym compounds manh-khoe da ban luoi-bieng toi-löi keu-goi

strong + strong' dirty + dirty' lazy + lazy' offense + fault' to call + to call'

'well in health' 'filthy' 'slothful' 'sin' 'to call upon, appeal'

Writing about a comparable construction in Hindi, Singh (1982) argues for a reduplication analysis in which a noun is repeated, but in different morphological forms. This is precisely the MDT view of such constructions. If a reduplication construction can constrain its daughters to be morphosemantically identical, certainly other constructions could constrain daughters to be near-identical or even to have opposing feature values (antonyms). MDT can relate reduplication to these constructions; phonological copying theories cannot. The parallelism is clear in languages like Acehnese. As documented by Durie (1985: 39-44), Acehnese possesses total reduplication (a), partial reduplication (b), a synonym compounding (c) - as well as an antonym construction in which semantic opposites are juxtaposed (d). Constructions (a-c) have the semantics of emphasis; juxtaposition of opposites (d) has a meaning that covers the meanings of both parts:

72 (8)

Sharon Inkelas a.

b.

Total reduplication tambö-tambö ma-ma tuleueng-tuleueng jamee-jamee

'drum-drum' 'mother-mother' 'bone-bone' 'guest-guest'

Partial reduplication singöh-ngöh (cf. singöh bube-be (cf. bube

'sometime indefinite in the future' 'tomorrow') 'as big as' 'size')

Suppletive allomorph/synonym constructions: irang-iröt 'zig-zag' (cf. irang, iröt 'skew') kreh-kroh 'rustling dry sound' (cf. kreh, kroh 'rustling dry sound') Juxtaposition of opposites tuha-muda 'old and young' bloe-publoe 'buy and sell' uroe-malam 'day and night' beungöh-seupöt 'morning and evening' These four constructions are united not just by the semantic correspondence they exhibit between the sister constituents, but also by phonology: all four construction types exhibit stress on each constituent, thus two stresses overall.1 From Durie (1985: 43-44): (9)

ureueng'-ureueng 1 geunap'-nap' lakoe'-binoe'

'people' (literally: 'person-person') 'every single one' 'men and women'

Normally, Acehnese allows just one, word-final stress per word. In Acehnese it is possible to establish a meta-construction, as shown below, which unites the synonym, antonym and reduplication constructions under one umbrella. (On meta-constructions, see e.g. Stump 1998 and references therein, or, for a near-equivalent, Bochner 1993.)

Morphological Doubling Theory (10)

73

Meta-construction underlying the Acehnese constructs in (9): Semantics: X + Y Phonology: no stress reduction (preserve input stress)

Semantics: X Phonology: final stress

Semantics: Y = ±X Phonology: final stress

Expressing the unity of the constructions in (9) is straightforward if reduplication is analyzed as morphological doubling; unity does not emerge, to the same extent, if reduplication is analyzed via a RED affix which phonologically copies the base.

8.

Evidence that identity in reduplication is morphosemantic

We have seen thus far that morphological constructions can require a semantic parallelism (identity, near-identity or anonymy) of their daughters, and that reduplication in at least some cases must target morphological, rather than prosodic, subconstituents of the base (see also work by Downing, e.g. Downing 1999a/b, 2000 on morphological factors in reduplication). Together these findings predict the existence of reduplication constructions which require semantic identity between their morphological daughters, but which allow all other kinds of identity to vary. In a sense we have already seen evidence for this prediction, insofar as synonym compounding in Khmer, Vietnamese and Acehnese is related to reduplication. In this section we examine more dramatic cases in which the daughters in reduplication can differ phonologically and morphologically as long as they mean the same thing. There are two circumstances under which this can occur: Base and reduplicant contain different suppletive allomorphs of same morpheme Reduplicant contains morphemes that the base does not (possible, in MDT, only when the morphemes in question do not cause a semantic discrepancy between base and reduplicant). We illustrate the first type of case with Sye (§8.1), and the second with Ndebele (§8.2).

74

Sharon

Inkelas

8.1. Suppletive allomorphy A pattern of intensifying stem reduplication in Sye (Erromangan; Crowley 1998) interacts with a pattern of suppletive stem allomorphy conditioned by morphological context. Most verb roots in Sye have two alternants, which for present purposes we may call Steml and Stem2. Stem2 forms appear in certain verb tenses and after echo subject markers; Steml forms appear elsewhere, and can be understood as the default stem type. We assume that each affix potentially selects for stem type. While historically the result of prefixal or left-edge phonological modifications, synchronically stem allomorphy appears to be suppletive in many if not all cases. Steml eve ah evinte evsor ocep omol oruc ovoli ovyuvag

Stem2 ampcah avinte amsor agkep amol anduc ampoli avyuampag

gloss 'defecate' 'look after' 'wake up' 'fly' 'fall' 'bathe' 'turn it' (causative prefix) 'eat'

Intensifying reduplication doubles the verb root. As shown in (12), when a reduplicated root occurs with a prefix selecting for a Stem2 root, the first copy is Stem2, but the second remains in default Steml form: (12)

a. Reduplication of stem 'fall' in Steml context: Steml-Steml

omol-omol

'fall all over'

b. Reduplication of stem 'fall' in "modified" stem context: Pfx-Stem2-Steml

cw-amol-omol

'they will fall all over'

Many interesting issues arise in constructions like that in (12b), which we will not be able to explore here. The example suffices to show that the reduplicants are not always phonological copies of their bases. What we are seeing in (12b) is semantic agreement between suppletive allomorphs: it is morphological doubling.

Morphological Doubling Theory

75

8.2. Empty morphs in reduplicant which are not in base Like many Bantu languages, Ndebele (Hyman et al. to appear, Sibanda 2004) possesses verb stem reduplication meaning 'do here and there, a little bit'. The reduplicant, which precedes the verb stem, is limited to two syllables, as shown in (13a). Stem-final inflectional suffixes do not reduplicate; only the root plus any following derivational suffixes (what Downing has called the Derivational Stem) is subject to reduplication. As shown in (13b), some stems are disyllabic only by virtue of containing an inflectional suffix. In such cases, regardless of whether the inflectional suffix is subjective -e, perfective -He, or default semantically empty -a, an [a] is provided as the second syllable of the reduplicant. We assume that this [a] can be equated with the semantically empty default suffix -a with which verbs must end when they do not end in a subjective or perfective suffix, and that the reduplicants in (13b) contain morphs not present in the 'base'. In (13), reduplicated stems are shown in the infinitive; unreduplicated stems, with suffixes set off by hyphens, are shown to the right. (13)

a.

uku-nambi-nambitha

b.

uku-lima-lima uku-lima-lime uku-lima-limi uku-lima-limile

'to 'to 'to 'to

'to taste'

/nambith-a/

cultivate' /lim-a/ cultivate, subjunctive' /lim-e/ cultivate, negative' /lim-i/ cultivate, perfective' /lim-ile/

If the verb stem is CV, where the V represents an inflectional suffix, even more radical discrepancies between reduplicant and base result. As shown in (14a), the reduplicant in such cases contains not only the empty morph -a but another semantically empty morph, [yi]. If, as shown in (14b), the stem has a consonantal root and any derivational suffixes, the reduplicant can freely choose among the derivational suffixes, empty -a or -yi, or any combination thereof that brings the reduplicant up to two syllables: (14)

a.

uku-dla^-dla uku-mayi-ma

'to eat' 'to stand'

/dl-a1 /m-a/

b.

uku-dlela-dlela uku-dleyi-dlela uku-dlavi-dlela

'to eat (applicative)' (ditto) (ditto)

/dl-el-a/

76

Sharon Inkelas

Like empty -a, -yi is used independently in the grammar; it is added (as a prefix) to roots in the imperative when they would otherwise be subminimal. Ndebele requires not only its reduplicants but also its verbs to be minimally disyllabic; imperatives consists of unprefixed stems, such that the imperative of a stem like dl-a 'eat' is rendered as yi-dl-a 'eat!'. 2 In conclusion, Ndebele permits reduplicants to contain not just one but even two semantically empty morphs not found in the following stem, which in BRCT would have to be the phonological base of reduplication. But Ndebele reduplicants are not phonological copies of the following stem; they are independently generated stems with the same meaning that bear the additional requirement of being truncated or augmented to two syllables. Morphotactic discrepancies of the kind seen in Sye and Ndebele are serious problems for BRCT, either falsifying the theory outright or requiring serious departures from its essential architecture, in which base and reduplicant are generated from a single input (see e.g. Downing 2000, who draws on semantically related words in generating the reduplicant). The facts fit naturally into a morphological doubling account in which the only kind of forced identity is semantic. MDT is not only compatible with the data; it predicts effects like those in Sye and Ndebele to occur.

9. Parallel phenomena in syntax In this section we explore a set of facts from syntax showing that constructions much like those in Sye and Ndebele are paralleled in syntax, where they have been called 'syntactic reduplication', but where a morpheme RED would be completely inappropriate.3 Here we discuss phrasal phenomena that are directly comparable to the morphological reduplication patterns in Sye and Ndebele.

9.1. Fongbe What Lefebvre and Brousseau (2002) call syntactic doubling in Fongbe (Kwa) occurs in four syntactic constructions: temporal adverbials (a), causal adverbials (b), factives (c) and predicate clefts (d). In each case a verb is doubled, with the extra copy appearing initially in the verb phrase.

Morphological Doubling Theory

77

The fronted copy of the verb can be identical to the main verb, or, for some speakers, it can also occur truncated to its first syllable (Collins 1994, cited in Lefebvre and Brousseau 2002: 505): (15)

a.

s i s o ~ s i Koku siso tlolo bo xesi tremble Koku tremble as.soon.as and fear 'As soon as Koku trembled, Bayi got frightened'

b.

siso ~ si Koku siso litu xssi tremble Koku tremble cause fear 'Because Koku trembled, Bayi got frightened'

c.

siso ~ si c|e-e Bäyi siso 5, νέ nu tremble OP-RES Bayi tremble, DEF bother for 'The fact that Bayi trembled bothered m e '

d.

siso ~ si we, Koku siso tremble it. is Koku tremble 'It is tremble that Koku did'

φ get

φ Bäyi get Bayi Bäyi Bayi mi me

In Fongbe, the only difference between the main verb and its fronted copy is the optional truncation; both copies can be assumed to have phonologically and semantically identical inputs. In Lango, however, we find a doubling construction which more resembles Sye or Ndebele morphological doubling in that the two copies can differ morphotactically and phonologically; agreement is semantic only.

9.2. Lango As described by (Noonan 1992: 175), Lango (Lwo, Western Nilotic) has an emphatic syntactic construction which repeats the verb. The first copy of the verb is inflected normally. The second copy, however, appears in what is called the gerund form; it "is given a high tone and preceded by ä- and followed by - ä . . . " : (16)

a.

äbino äbino 1 SG.COME.PERF come.GER Ί did come yesterday'

b.

jiäkö omyelö girl 3SG.dance.PERF 'the girl just danced'

äwo'ro yesterday

ämyelä dance.GER

78

Sharon

c.

Inkelas

Iocs

öneko

gwök

man 3SG.kill.PERF dog ' t h e m a n killed the d o g '

äneks

okkd

kill.GER

completely

The gerund form of the verb semantically has a subset of the features of its inflected counterpart; the construction resembles Sye and Ndebele morphological reduplication in supplying morphotactically different, but semantically similar, copies of the doubled verb.

9.3. Chechen (and Ingush) Chechen (North Caucasian) exhibits syntactic reduplication to satisfy the requirements of a second position clitic (Conathan and Good 2000; see also Peterson 2001 on Ingush). As shown in (17), chained clauses are marked by an enclitic particle Pa, which immediately precedes the inflected, phrasefinal, main verb. The enclitic must be preceded by another element in the same clause. Two types of constituent may occur before the verb (and enclitic particle) in the clause: an object (a), or a deictic proclitic or preverb (b). If neither of these elements is present in a chained clause, then the obligatory pre-clitic position is filled by reduplicating the verb (c): (17)

a.

Cickuo,

[ch?aara = ?a gina]vp,

cat.ERG

[fish

=&

?i

see.PP]Vp 3S.ABS

bu?u B.eat.PRES

'The cat, having seen a fish, eats it' b.

Ahmad, [kiehkat jaaz =?a dina]Vp, zhejna dueshu Ahmad.ERG [letter write =& D.do.PP]Vp book D.read.PRS 'Ahmad, having written a letter, reads a book'

c.

Ahmad, [?a =?a ?iina] V p, Ahmad [stay.RED =& stay.PP] VP 'Ahmad stayed (for a while) and left'

dia-vaghara DX.V.go.WP

The Chechen reduplicant occurs in infinitive form, while the main verb is inflected. Because inflection sometimes requires use of a suppletive form of the root (cf. two vs. Dala for 'give'), Chechen can exhibit Sye-like suppletive allomorphy differences between base and reduplicant. It is a virtue of MDT that the kinds of structures it posits for morphological reduplication - a constituent containing two semantically identical morphemes - are, however they are generated, the same kinds of structures that are needed to describe syntactic reduplication. It is clear that what is

Morphological Doubling Theory

79

going on in (17c) is the use of two verbs with (almost) the same meaning; it would be absurd to say that what precedes the clause-chaining clitic is a reduplicative morpheme RED, which in any case is not a phonological copy of the main verb.

10. Where MDT draws the line: phonological duplication We have emphasized that in MDT the defining property of reduplication is semantic, rather than phonological identity. There are, to be sure, phenomena that have been called reduplicative, in that a phonological element is doubled, but which are not amenable to a morphological doubling analysis, in part because the doubled element is something very small, like a single consonant or vowel, and in part because the doubling has a purely phonological purpose, rather than being associated with a change in meaning. For example, Hausa (Chadic) has numerous noun pluralization constructions, the most productive of which involves a suffix whose medial consonant is a copy of the final consonant of the noun stem (Newman 2000: 431-32): (18)

Hausa productive noun pluralization bindigä: fanni: hüku:mä:

bindig-o:gi: fann-o:ni: huku:m-o:mi:

'gun' 'category' 'governmental body'

In Spokane (e.g. Black 1996: 210 ff., Bates and Carlson 1998), the repetitive form of a verb is formed by infixing /e/ into an initial consonant cluster, if any (19a); for verbs beginning with only a single consonant, that consonant is doubled, with /e/ appearing between the two copies (19b): (19)

Spokane repetitive infixation a.

b.

/-e-, sal'-n'-t-anV

s-e-l'n'ten'

/REP, chop-CTR-TR-1 SGTRS/

Ί cut it up repeatedly'

w

/-e-, nic'-n'-t-3x / REP, Cut-CTR-TR-2SGTRS

-*· η -e-n ic η tx. w 'you kept cutting'

In both Hausa and Spokane, the duplication of the consonant is driven purely phonologically, by the need for a syllable onset. Autosegmental phonology would spread a consonant to the onset position; in Optimality The-

80

Sharon Inkelas

ory ONSET could compel the insertion of a consonant which agrees featurally with a nearby consonant. Walker (2000), Hansson (2001), Rose and Walker (2004) and Zuraw (2002) have developed theories of string-internal agreement which automate string-internal correspondence among segments, assuming they meet certain thresholds of similarity and locality; if FAITHio is ranked low, corresponding output segments can be made to become more (or less) similar along additional dimensions of similarity. 4 (Although for lack of space we cannot work out the details here, the approach will favor copying (e.g., from (19b), n-e-nic'n'tax") over epenthesis of a default consonant (e.g. P-e-nic'n 'tax ") if correspondence and identity requirements are ranked higher than segmental unmarkedness.) It might appear that we now have two methods for duplicating material: morphological reduplication, which we analyze in terms of doubling, and phonological duplication, which we analyze as phonological spreading or string-internal agreement. Hendricks (1999) and Gafos (1998) have both noted that theories with both morphological reduplication and phonological spreading are potentially redundant, insofar as both methods are applicable to the same data. Hendricks and Gafos propose to eliminate spreading in favor of morphological reduplication. However, this approach presupposes that there are data which could be analyzed in both ways. On the contrary, we argue, the phenomena for which morphological doubling is appropriate are very different from those for which phonological doubling is appropriate; both approaches should be retained, with no overlap. Below we list some criteria for determining when a copying effect is reduplication and when it is phonological duplication. (1) Phonological duplication serves a phonological purpose; morphological reduplication serves a morphological process (either by being a word-formation process itself or by enabling another wordformation process to take place; see e.g. the discussion in Inkelas and Zoll (to appear) of Nancowry, in which reduplication of monosyllabic roots is the means of satisfying a disyllabic stem condition imposed by a particular affix). (2) Phonological duplication involves a single phonological segment, as in Hausa or Spokane onset-driven consonant copying; morphological reduplication involves an entire morphological constituent (affix, root, stem, word), potentially truncated to a prosodic constituent (mora, syllable, foot)

Morphological Doubling Theory

81

(3) Phonological duplication involves, by definition, phonological identity, while morphological reduplication involves semantic, not necessarily phonological, identity (4) Phonological duplication is local (a copied consonant is a copy of the closest consonant, for example), while morphological reduplication is not necessarily local. We have seen cases in syntactic reduplication in which the two copies are separated by other words; many parallel examples, in which base and reduplicant are separated, exist in morphology as well, e.g. Chukchee nute-nut 'earth (absolutive singular)' (Krause 1980), Umpila maka 'die, go out' -» maka-l-ma 'die, go out (progressive)' (Harris and O'Grady 1976, Levin 1985).

10.1. Phonological correspondence and backcopying Although phonological duplication is appropriate only for phonological, not morphological, duplication, in involving string-internal correspondence it formally resembles the BRCT approach to morphological reduplication. This fact causes BRCT and MDT to generate rather different predictions about the phonology of reduplication. We explore one of these here. Unlike MDT, BRCT assumes that reduplicant and base are in bidirectional phonological correspondence, making it possible not only for the base to influence the reduplicant, but for the reduplicant to influence the base as well. The latter phenomenon, anticipated in Wilbur 1973, is termed 'backcopying' by McCarthy & Prince (1995). Backcopying in morphological reduplication is a phenomenon which MDT cannot describe, because there is no sense in which the reduplicant ever corresponds to the base. By contrast, however, we do assume string-internal correspondence in our analysis of phonological duplication. Thus, we make the following prediction: (20)

Backcopying should occur with phonological duplication (assimilation), but not directly as a result of morphological reduplication

Our surveys have not found backcopying to be a robust feature of morphological reduplication. Consider the following hypothetical example, based on a discussion in McCarthy & Prince 1995: 326, in which the effects of nasal place assimilation across the reduplicant-base boundary are reflected in the base:

82 (21)

Sharon Inkelas Hypothetical backcopying junctural assimilation (RED = σ) a. b.

RED-kama RED-pana

-> kaq-kaqa pam-pama

Nasal place assimilation is one of the most common junctural phonological alternations; if backcopying is a real effect, we should expect to see cases like that in (21), but none, to our knowledge, have been documented. McCarthy & Prince (1995) do offer a number of suggestive examples of other kinds which they characterize as involving backcopying; however, most of these have turned out on closer inspection to have a rather different character; see e.g. Inkelas and Zoll (to appear), Zoll 2002, Raimy 2000. Backcopying thus does not appear to be a feature of morphological reduplication, supporting the MDT approach over the BRCT approach. But backcopying does appear to occur in the kind of examples that we independently classify as phonological duplication, supporting the prediction in (21). We illustrate with one example from Hausa participle formation (Newman 2000: 19). Like the plural formation construction discussed earlier, the participle formation construction involves a disyllabic suffix, -aCCe:, whose medial consonant is supplied through duplication of the final consonant of the stem (22a). The second syllable of the participial suffix has a front vowel, [e:]. Hausa palatalizes coronals before front vowels; a duplicated stem-final coronal will palatalize internal to the participial suffix. As noted in McCarthy 1986, Newman 2000: 417, this palatalization effect is occasionally extended to the stem-final consonant, resulting in the overapplication of palatalization (22c). Anticipation of palatalization is a backcopying effect: (22)

Hausa participle formation: X Q V -*· XC-aCiQe: a. b. c.

dafä: tfikä: fasä: fasä:

däf-affe: tjik-akke: fäs-ajje: fäj-ajje:

'cook/cooked' 'fill/full' 'break/broken' 'break/broken' (sporadic)

Compare this actual case to a morphological reduplication process in Hausa that does not exhibit backcopying. As shown in (23), Hausa forms pluractional verbs through morphological reduplication; a verb is doubled and the first copy truncated to its initial CVC portion, which surfaces as a closed syllable. If the stem-initial vowel is long, it must shorten in the reduplicant,

Morphological Doubling Theory

83

since Hausa does not permit long vowels in closed syllables. Reduplicantfinal noncoronal obstruents assimilate totally to the following consonant, producing a geminate consonant at the base-reduplicant juncture, as in (23a) (Newman 2000: 424-25). Gemination is also an option for reduplicant-final sonorants and coronal obstruents (23b); however, for these categories of segments, gemination is not obligatory. (23c-e) show the nongemination options for reduplicant final sonorants and coronal obstruents (both variants are provided for 'go out'). Reduplicant-final sonorants surface intact, modulo nasal place assimilation (23d) and glide vocalization (not shown); reduplicant-final coronal obstruents can also resist gemination (23e), though they do sonorize and rhotacize. Plain stem

Reduplicated stem

Form with backcopying

a.

'open mouth widely' 'step on'

wa:ge: ta:ka:

waw-wä:ge: tat-tä:kä:

*waw-wäwe: *tat-täta:

b.

'oppress' 'go out' 'sell'

danne: fitä: sayar

dad-dänne: fif-fitä: sas-sayar

*däd-dädde: *fif-fifa: *sas-sasar

c.

'chip off'

ßalgäta:

ßal-ßalgäta:

d.

'catch'

ka:mä:

kaq-kä:ma:

*ka:-kä:a:

e.

'go out' 'kill'

fitä: kaje:

firfitä: kar-kaje:

*fir-fira: *kar-kare:



Neither gemination (23a-b), nor nasal place assimilation (23d), nor coronal sonorization (23e), nor closed syllable vowel shortening, attested in reduplicants throughout (23), is ever backcopied to the base, nor would we expect such an effect in Hausa or any other language. The prediction in (20) thus appears accurate. The fact that backcopying occurs in phonological assimilation but not in morphological reduplication per se strongly supports not only the use of MDT for morphological reduplication but also the conclusion that phonological duplication and morphological reduplication are entirely different processes.

84

Sharon Inkelas

11. Conclusion This paper has argued that semantic identity is what defines the two copies in reduplication constructions; phonological and morphotactic identity, while common correlates of semantic identity, are not required, and indeed we find variation along both phonological and morphotactic dimensions. In modeling reduplication as morphological doubling, rather than phonological copying, M D T thus achieves greater descriptive adequacy than phonological copying theories, such as BRCT. At the same time M D T is more restrictive theoretically than BRCT in particular, which overgenerates by predicting unattested backcopying phenomena. By seeing reduplication as an essentially morphological process, M D T is better able to model the facts, as well as to relate reduplication to the other sister constructions that we have documented, both in morphology and in syntax.

Notes *

1.

2.

3.

Morphological Doubling Theory is the result of joint work with Cheryl Zoll; a co-authored, booklength presentation of the theory and results is currently underway (Inkelas & Zoll, forthcoming). I am grateful to an anonymous reviewer for comments on an earlier draft of the paper. Durie notes, however (p. 43), that "[f]or words of three syllables or more the emphatic semantic effect of doubling is achieved by simply reduplicating the initial syllable. The resulting construction has only a single word stress." One might be tempted to analyze either a or yi as phonologically epenthetic. However, such an approach would run into trouble with (a) the fact that a occurs verb-fmally, when no other inflectional suffix is appropriate, to fill the final obligatory suffix position, even when the preceding morpheme is vowelfinal and epenthesis could not be motivated (e.g. /bal-u-a/ -» [balwa] 'readPASS-a', (b) the fact that yi is a cross-linguistically marked syllable and not what one would expect as the result of epenthesis, and (c) the fact that yi appears either in its entirety or not at all; if yi were the result of phonological epenthesis of unmarked material we would also expect to find, in (14b), the following to be grammatical reduplicants: *dleli, from /dl-el-/ and i epenthesis; or *dleya, either from /dl-e/ and epenthesis of a and y, or from /dl-e-a/ and y epenthesis. These reduplicants are not possible. For a discussion of reduplication of syntactic phrases, or, rather, the phonological phrases derived from syntactic structure, see Cole 1994.

Morphological 4.

Doubling Theory

85

While the mechanisms they use are fundamentally similar, Zuraw characterizes the resulting string-internal correspondence as reduplication, while Hansson, Rose and Walker characterize it as harmony (all three discuss primarily consonant harmony).

References Applegate, Richard 1998 Spokane syllable structure and reduplication. In Salish languages and linguistics: theoretical and descriptive perspectives, Ewa Czaykowska-Higgins and Dale Kinkade (eds.), 99-123. (Trends in Linguistics: Studies and Monographs 107). Berlin/New York: Mouton de Gruyter. Black, Deirdre 1996 The morphological and phonological structure of Spokane lexemes. Ph.D. dissertation, University of Montana. Bochner, Harry 1993 Simplicity in generative morphology. Berlin/New York: Mouton de Gruyter. Cole, Jennifer 1994 A prosodic theory of reduplication. Ph.D. dissertation, Stanford University. Collins, Chris 1994 The factive construction in Kwa. Travaux de recherche sur le Creole hai'tien 23: 31-65. Conathan, Lisa, and Jeffrey Good 2000 Morphosyntactic reduplication in Chechen and Ingush. In Proceedings of the Chicago Linguistic Society 36-2: the panels, Arike Okrent and John P. Boyle (eds.), 49-61. Chicago: Chicago Linguistic Society. Crowley, Terry 1998 An erromangan (Sye) grammar. (Oceanic linguistics special publication ; no. 27). Honolulu: University of Hawaii Press. Dixon, Robert 1972 The Dyirbal language of North Queensland. Cambridge: Cambridge University Press. Downing, Laura 1999a Morphological constraints on Bantu reduplication. Linguistic Analysis 29: 6-46. 1999b Verbal reduplication in three Bantu languages. In The prosodymorphology interface, Harry van der Hülst, Rene Kager and Wim Zonneveld (eds.), 62-89. Cambridge: Cambridge University Press.

86

Sharon 2000

Inkelas Morphological and prosodic constraints on Kinande verbal reduplication. Phonology 17: 1-38.

Durie, Mark 1985 A grammar of Acehnese: on the basis of a dialect of North Aceh. Dordrecht: Foris. Gafos, Diamandis 1998 Eliminating long-distance consonantal spreading. Natural Language and Linguistic Theory 16: 223-278. Hansson, Gunnar 2001 Theoretical and typological issues in consonant harmony. Ph.D. dissertation, University of California, Berkeley. Harris, B. P., and G. N. O'Grady 1976 An analysis of the progressive morpheme in Umpila verbs. In Languages of Cape York, P. Sutton (ed.), 165-212. Canberra: Australian Institute of Aboriginal Studies. Harrison, Sheldon P. 1973 Reduplication in Micronesian languages. Oceanic Linguistics 12: 407^154. 1976 Mokilese Reference grammar. Honolulu: The University Press of Hawaii. Hendricks, Sean Q. 1999 Reduplication without template constraints: a study in bareconsonant reduplication. PhD dissertation, University of Arizona. Hyman, Larry, Sharon Inkelas, and Galen Sibanda forthc. A subtree correspondence theory of reduplication in Ndebele. In The Nature of the Word: Essays in Honor of Paul Kiparsky, Sharon Inkelas (ed.). Cambridge: MIT Press. Inkelas, Sharon, and Cheryl Zoll forthc. Reduplication: Doubling in Morphology. Cambridge University Press. Krause, Scott 1980 Topics in Chukchee Phonology and Morphology. PhD dissertation, University of Illinois Urbana-Champaign. Lefebvre, Claire, and A. M. Brousseau 2002 A grammar of Fongbe. Berlin/New York: Mouton de Gruyter. Levin, Juliette 1985 Reduplication in Umpila. In MIT Working Papers in Linguistics, vol. 6: Papers in Theoretical and Applied Linguistics, Diana Archangeli, Andrew Barrss and Richard Sproat (eds.). Cambridge: Massachusetts Institute of Technology. Marantz, Alec 1982 Re reduplication. Linguistic Inquiry 13: 483-545.

Morphological

Doubling Theory

87

McCarthy, John 1986 OCP effects: gemination and antigemination. Linguistic Inquiry 17: 207-263. McCarthy, John, and Alan Prince 1993 Prosodic morphology I: constraint interaction and satisfaction. Unpublished manuscript, University of Massachusetts, Amherst and Rutgers University. 1995 Faithfulness and reduplicative identity. In University of Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory, Jill Beckman, Laura Dickey and Suzanne Urbanczyk (eds.), 249-384. Amherst, MA: GLSA. 1999 Prosodic Morphology (1986). In Phonological theory: the essential readings, John Goldsmith (ed.), 238-288. Maiden, MA: Blackwell. Newman, Paul 2000 The Hausa language: an encyclopedic reference grammar. New Haven: Yale University Press. Nguyen, Dinh-hoä 1997 Vietnamese: Tieng Viet Khong Son Phan. Amsterdam: John Benjamins. Noonan, Michael 1992 A grammar of Lango. Berlin/New York: Mouton de Gruyter. Oura, Noeurng, and John Haiman 2000 Symmetrical compounds in Khmer. Studies in Language 24: 4 8 3 514. Peterson, David 2001 The Ingush clitic ?a: the elusive type 5 clitic? Language 77: 144155. Raimy, Eric 2000 The phonology and morphology of reduplication. Berlin/New York: Mouton de Gruyter. Rice, Keren 1997 Japanese NC clusters and the redundancy of nasal voicing. Linguistic Inquiry 28: 541-551. Roberts, John 1987 Amele. London: Croom Helm. 1991 Reduplication in Amele. Papers in Papuan Linguistics 1: 115-146. Rose, Sharon, and Rachel Walker 2004 A typology of consonant agreement as correspondence. Language 80: 475-531. Sibanda, Galen 2004 Verbal phonology and morphology of Ndebele. PhD dissertation, University of California, Berkeley.

88

Sharon

Inkelas

Singh, Rajendra 1982 On some 'redundant compounds' in Modern Hindi. Lingua 56: 3 4 5 351. Steriade, Donca 1988 Reduplication and syllable transfer in Sanskrit. Phonology 5: 7 3 155. Stump, Gregory 1998 Comments on Inkelas and Orgun's paper. In Morphology and its interaction with syntax and phonology, Steven Lapointe, Diane Brentari and Patrick Farrell (eds.), 3 9 3 ^ 0 5 . Stanford, CA: CSLI Publications. Walker, Rachel 2000 Long-distance consonantal identity effects. In Proceedings of the 19'h West Coast Conference on Formal Linguistics, Roger Billerey and Brook Lillehaugen (eds.), 532-545. Somerville: Cascadilla Press. Wilbur, Ronnie 1973 The phonology of reduplication. Bloomington: Indiana University Linguistics Club. Zoll, Cheryl 2002 Vowel reduction and reduplication in Klamath. Linguistic Inquiry 33:520-527. Zuraw, Kie Ross 2002 Agressive reduplication. Phonology 19: 395-439.

The Emergence of the Marked: Tone in some African reduplicative systems1 Laura J. Downing

1.

Introduction

The defining hallmark of the process of reduplication is that the Base and the reduplicated portion (RED) of the word have identical phonological substance (Wilbur 1973; Steriade 1988; McCarthy & Prince 1993, 1995). Deviations from perfect identity are common, however, and typically involve reduction in the size or featural markedness of the RED compared to the Base (Alderete et al. 1999; Niepokuj 1991). As work like McCarthy & Prince (1995, 1999) shows, this can be explained within Optimality Theory by two morphologically-based asymmetries. First, the Base and RED have an asymmetrical phonological relation to the input. The Base - arguably the head of the reduplicative construction - has a direct relation (through InputOutput correspondence) - to an input segmental string. The RED has an indirect relationship to an input segmental string, via the Base (Base-RED correspondence). 2 As Alderete et al. (1999) show, this asymmetrical relationship allows constraint ranking to account for why the RED often has less marked phonological structure than the Base: what they term The Emergence of The Unmarked (TETU) effect. A second asymmetry is defined by the morphological category assigned to RED. If RED is a major morphological category like Word or Stem, forming a compound with the Base, it is predicted to meet the usual (bisyllabic or bimoraic) minimality requirement on these categories (Niepokuj 1991). If RED is a bound category like Root or Affix, it can be subminimal. Further, as Urbanczyk (1996, 2000) argues, Affixal REDs are expected to contain less marked prosodic and featural structure, in keeping with other Root-Affix asymmetries (see, e.g. Beckman 1998). These two asymmetries account for many tonal mismatches between the RED and Base. The RED often has the unmarked tone (Akinlabi 1997; Alderete et al. 1999), while the Base has the marked tone. And it falls out that the RED has the unmarked tone, if it has the morphological status of

90

Laura J. Downing

Affix, rather than Word, Stem or Root (Myers & Carleton 1996; Hyman & Mtenje 1999). These points will be illustrated in the next section. However, as I will show in section 3, many languages show the opposite tonal asymmetry: the RED often has a marked tone, while the corresponding Base element has the unmarked tone. I will argue in section 4 that this actually follows from two autosegmental properties of tone. Morphemes can be specified for a tone (or a tone melody) independent of their segmental content. And tones associate to positions over some larger morphophonological domain, often outside the morpheme which contributes them to the input.

2.

The Emergence of the Unmarked

Current models of reduplication predict that, when RED and Base tone do not match, the tone of the RED should be unmarked compared to that of the Base, especially if RED is an Affix. This predicted pattern is attested in West African languages like Ewe/Gbe (Akinlabi 1997; Capo 1991) and Nupe (Akinlabi 1997; Smith 1969). For example, in Nupe, a Benue-Congo language spoken in Nigeria, the gerundive is formed by reduplicating the Base verb. As shown in (1), RED is always a single CV syllable (no matter how long the Base is) with a fixed high vowel (no matter what height the corresponding Base vowel is) and a Mid tone (no matter the tone of the corresponding Base vowel is). (1) (a)

Nupe gerundive reduplication (Akinlabi 1997; Smith 1969) Monosyllabic verbs gi be kpä do tswä

(b)

'eat' 'come' 'drizzle' 'praise' 'take care'

gi-gi bi-be kpi-kpä du-do/di-do tsu-tswä

'eating' 'coming' 'drizzling' 'praising' 'care'

ji-jakpe gi-gäya gu-goba ku-kiita

'stooping' 'being too long' 'surrounding' 'overlapping'

Polysyllabic verbs jäkpe gäya goba kuta

'stoop' 'be too long' 'surround' 'overlap'

The emergence of the marked

91

As Akinlabi (1997) argues, the fixed [+high] vowel and Mid tone of the RED are predicted TETU effects, as [+high] and Mid tone are the unmarked values for these features. 3 This phonological reduction is expected since RED is clearly an affix: it changes the lexical category of the Base from a verb to a noun. Morphological reduction and tone reduction do not always go together, however, as we find languages where the only asymmetry between RED and the Base is unmarked tone on one half of the reduplicative complex. For example, in Yao, a Bantu language spoken in Malawi, the entire verb stem is reduplicated to give the meaning of doing the action of the verb repeatedly. As shown in (2), the stem High tone is realized only on the first half of the reduplicative complex. The second half has only Low tones, arguably the unmarked tone in a two-tone system like that of Yao (Pulleyblank 1986): (2)

Yao verbal reduplication (Myers & Carleton 1996; Mtenje 2002; ku- is the infinitive prefix; the verb stem follows this prefix) (a) (b) (c)

Infinitive ku-teleka ku-womboka ku-sülümunda

Gloss 'to cook' 'to save' 'to sift (flour)'

to X repeatedly ku-teleka-teleka ku-womboka-womboka ku-sulumunda-sulumunda

Similarly, Awoyale (1989) shows that intensive ideophones in Yoruba, a Benue-Congo language spoken in Nigeria, are formed by total reduplication of the Base. As we see in (3), only one half of the ideophone bears a marked (High or Low) tone. The other half obligatorily bears the unmarked Mid tone: (3)

Yoruba intensive ideophones (Awoyale 1989) (a) (b) (c) (d)

yüngbä-yungba gbagidi-gbagidi rübütü-tü-rubutu-tu dodo-dodo

(yüngbä) (gbagidi) (rübütu) (dodo)

'honey-sweet' 'very fleshy and bulky' 'very cute and robust' 'severely sick or run down'

To account for cases like these, Myers & Carleton (1996) propose that the RED actually has the morphological status of Affix, so that tonal reduction follows from assigning RED the status of a morpheme that cannot bear contrastive tone. 4 However, there are problems with this hypothesis, as ar-

92

Laura J. Downing

gued in Downing (2003). Total reduplication clearly meets Inkelas & Zoll's (2000; Inkelas, this volume) definition of reduplication as selfcompounding at the phonological and morphological level. Our expectation is for each member of the reduplicative complex to have the same morphological category: verb Stem for Yao, Word for Yoruba. Further, as Urbanczyk (1996, 2000) argues, cross-linguistically Affixal REDs - like other affixes - tend to be monosyllabic, while longer REDs are root or stem-like. Indeed, in Yao and Yoruba, other affixes are generally monosyllabic. The REDs found in total reduplication would be a striking exception to this generalization about affix size. As a result, it is not plausible to propose that RED is an affix to explain the lack of tonal identity in languages like Yao and Yoruba where we find total reduplication of long Base morphemes. To sum up this section, tonal asymmetries in reduplication often provide evidence of the sorts of TETU effects predicted by the asymmetrical relationship of the Base and the RED to the input defined by Base-RED correspondence. There are numerous examples of languages where only one half of the reduplicative complex bears a marked tone, while the other half bears an unmarked tone. As we have seen, though, tonal asymmetries do not always fall out from the morphological category assigned to RED. Tonal neutralization is frequently found even when RED is segmentally identical to a Stem or Word Base and arguably has the same morphological category. Phonological reduction does not always correlate, then, with morphological reduction of the RED compared to the Base.

3.

The Emergence of the Marked

Both morphologically-based explanations for tonal mismatches rest on the assumption that RED can only get marked tonal features by copying them from the corresponding mora of the Base. In fact, the following brief survey of tone and reduplication in some African languages shows that there are two other important sources of RED tone that can lead to a tonal asymmetry between the RED and the Base. First, RED can contribute a marked tone (or tone melody) independent of the tone of the Base. Second, a marked tone contributed by the Base can be realized directly on RED instead of the Base, when the reduplicative complex forms a single domain of Base tone (melody) realization. Both cases can lead to what I call the Emergence of the Marked: the RED has a marked tone different from that of the corresponding Base string or the related unreduplicated Base form. 5

The emergence of the marked 3.1.

93

RED contributes marked tone (melody)

One source of marked tone on the RED that can lead to tonal asymmetry with the Base is that RED, like other affixes, can be lexically specified for a marked tone (or tone melody). Two different tone patterns on reduplicative constructions are produced when RED contributes a marked tone (melody). One pattern is for the RED's marked tone to be realized within the RED, while the Base realizes its independent input tone. The other pattern is for the RED's tone pattern to be dominant, realized over the entire RED+Base complex, so that the Base input tone is unrealized. Each of these is discussed in turn below.

3.1.1. RED marked tone realized within RED In some languages where the RED can contribute a marked tone, it is realized within the RED, leading to tonal asymmetry between RED and its Base if the Base has a different input tone. An example of this is provided by Yoruba gerundive reduplication (Alderete et al. 1999; Akinlabi 2000; Orie 1997; Pulleyblank 1988, to appear), illustrated in (4). Gerundives in Yoruba, as in Nupe (1), are formed by prefixing a CV RED with a fixed [+high] vowel to the Base verb. However, in Yoruba, unlike Nupe, the prefix has a fixed High tone, no matter what the tone of the Base is: (4)

Yoruba gerundives (Akinlabi 2000, Pulleyblank 1988) (a) (b) (c) (d) (e)

gbönä mu du lä je

'be warm' 'drink' 'scramble' 'split' 'eat'

gbi-gbönä mi-mu di-dü/dü-dü li-lä ji-je

'warmth; heat' 'drinking' 'scrambling' 'splitting' 'act of eating'

As Akinlabi (2000) and Pulleyblank (to appear) argue, Mid tone is the unmarked tone of Yoruba, so the fixed High tone on the RED cannot be analyzed in terms of TETU. Instead, the High tone associated with the RED in the output must be contributed by the input of the reduplicative construction. 6 This Emergence of the Marked effect is shown most strikingly in (4b,e), where we see the RED has a marked High tone even when the Base has the unmarked Mid tone.

94

Laura J. Downing

A similar tonal asymmetry is found in Kuku (Cohen 2000), a dialect of Bari (East Nilotic) spoken in Uganda. In this language there are two tone classes of verbs, Η and LHL. The past continuous form of verbs is formed by prefixing a High-toned CV RED to the verb. As shown in (5), when the following verb is in the LHL tone class, the Low tone of the verb's tone melody occurs on the first syllable of the verb stem. The corresponding RED syllable preceding the verb has its input High-tone, leading to a tonal asymmetry between the RED and Base: (5)

Kuku past continuous (Cohen 2000: 28); RED is underlined (a) (b) (c)

η xz_ re-ja do' so'-so-jQ ma'tat me'-med-da

Ί was sweeping' 'you were boiling something' 'the leader was looking'

In Kukü, as in Yoruba, High tone is the marked tone, so this tonal asymmetry cannot be a TETU effect but instead reveals the Emergence of the Marked. The RED has a fixed marked tone contributed by the reduplicative construction. These are not isolated examples. Ibibio (Akinlabi 1997; Akinlabi & Urua 2000) and Igbo (Clark 1990) are also described as having reduplication patterns in which RED has a fixed marked tone, independent of the Base tone.

3.1.2. Dominant RED tone In other languages, like Hausa (Newman 2000), Arusa Maasai (Levergood 1991) and Yoruba (Awoyale 1989, Orie 1997), RED-Base tone asymmetries arise when the tone (melody) contributed by RED dominates the input tone of the Base. 7 For example, Newman (2000, chapter 56) identifies 15 different major ways to form nominal (noun and adjective) plurals in Hausa, a Chadic language spoken mainly in Nigeria. Almost all introduce a tone melody which is realized over the entire plural nominal word, often replacing the tone of the related singular form. Many involve partial or total reduplication (rather than a fixed segment affix alone). In the examples in (6a) - (6d), for example, the plural is formed by reduplicating the final CVC of the singular stem, suffixing -i: following this reduplicative syllable, and introducing a HLHH dominant tone melody. Notice that other plu-

The emergence of the marked

95

rals in this class, like those in (6e)-(6g), which are not formed by reduplication, have parallel prosody: four syllables and a HLHH tone melody: (6)

Hausa plurals, class 8 (Newman 2000: 450-452); RED is underlined

(a) (b) (c) (d) (e) (f) (g)

Singular

Plural

Gloss of singular

gärdäm-ä ämär-yä: shä:wär-ä: kädan-yä: färs-ä: gärm-ä: sälk-ä:

gärdändäm-i: amärmar-i: sha:wärwär-i: kaäänäan-i: färe:s-äni: gäre:m-äni: säle:k-äni:

dispute, argument bride advice, counsel shea tree or nut split kola nut large hoe foot soldier

While there is a tonal asymmetry between the RED and the corresponding Base string in (6a) - (6d), this is clearly not a TETU effect. As we can see by comparing these forms with the plurals in (6e) - (6g), all plurals in this class have a Low tone on the second syllable, and the remaining syllables bear a High tone, no matter what the input tone of the singular stem might be. The RED has a different tone from the corresponding Base string, then, because it is in a position to be associated with a High tone, while the Base is in the position associated with the Low tone of the plural tone melody. The tone of RED does not depend - either through identity or markedness motivated dissimilation - on the tone of its corresponding Base string. Dominant RED tone is also characteristic of many Yoruba reduplicated ideophones (Awoyale 1989). As shown in (7), the reduplicated ideophone often has a different tone melody from the unreduplicated form given in parentheses: (7)

Yoruba ideophones (Awoyale 1989, Orie 1997) (a) (b) (c) (d) (e)

gbälä-gbälä yigi-yigi yigi-yigi jäla-jäla wuru-würu

(gbälä) (yigO (yigO (jälä) (würü)

'very free or loose movement' 'not walking straight/upright' 'not walking straight/upright' 'moving shabbily' 'disorderly'

Notice that the reduplicative tone melody is distributed over the entire reduplicative complex in these ideophones, replacing the original tone of the unreduplicated form. This leads to a tonal asymmetry between the two

96

Laura J. Downing

halves of the reduplicative complex, with marked (High and Low) tones found on both halves in many of the forms in (7). In Yoruba, as in Hausa, then, Emergence of the Marked results when the reduplicative construction introduces a tone melody containing marked tones which are distributed over the RED+Base complex in such a way that both halves are realized with mismatched (marked) tones.

3.1.3. Formal

analysis

I propose that both types of tonal asymmetries discussed in the preceding sections which result when RED contributes a marked input tone can be accounted for by properly ranking the constraints in (8): (8)

Constraints defining RED tone assignment (a)

DOMAFFIX

Align(AffixT, Word): Align the tone melody of the Affix within the entire word. (b)

ALIGNMORPH

Align(MorphT, Morph): Align the tone melody of a morpheme within the morpheme that contributes it to the Input. (c)

FAITH-BR(TONE):

For corresponding moras in the Base and RED, each mora of the Base should have the same tone in RED and each mora of RED should have the same tone in the Base. (d)

FAITH-IO(TONE):

For corresponding moras in the Input and Output, each mora of the Input should have the same tone in the Output and each mora of the Output should have the same tone in the Input. DOMAFFIX (8a) optimizes aligning the tone (melody) of the reduplicative affix within the entire reduplicative word. In the languages with dominant reduplicative tone described in section 3.1.2, above, this constraint must be highly ranked. ALIGNMORPH (8b) optimizes realizing the tone of each mor-

The emergence of the marked

97

pheme within the morpheme that contributes it to the input. This constraint must be highly ranked in the languages described in section 3.1.1 where the tone contributed by the RED is also realized within the RED. I am following work like McCarthy & Prince (1999) and Alderete et al. (1999) in using FAITH-BR (8C) and FAITH-ΙΟ (8d) as cover terms for the Faithfulness constraints optimizing tonal identity between RED and the Base or Input and Output, respectively. In both types of languages discussed in section 3.1, FAITH-BR (8C) is low-ranked, as the source of the tone on RED is never the Base. The tableau in (9) exemplifies the analysis of the languages where reduplicative tone is dominant, like the Hausa plurals in (6). (The analysis of the Yoruba ideophone data in (7) would be essentially identical, so it is not given here.) (9)

Hausa plurals (6)

gärdäm-RED-I:, HLHH

DOM AFFIX

gärdändäm-i: b. gärdändäm-i:

*!

ALIGN MORPH

FAITH10

FAITH -BR

*

*

*

*

Constraint ranking optimizing dominant affix (RED) tone: DOMAFFIX (8a) »

ALIGNMORPH (8b), FAITH-ΙΟ (8d), FAITH-BR (8c)

Candidate (9a), in which the HLHH reduplicative melody is realized over the entire reduplicative word, is optimal, as it satisfies the highest-ranked constraint, DOMAFFIX (8a). 8 The competing candidate is non-optimal. Even though it satisfies the correspondence constraints FAITH-ΙΟ and FAITH-BR - the moras of the Base have the same tone as in the input and the RED and Base have the same tone - it violates DOMAFFIX, as the plural tone melody is not realized over the output plural string. The tableau in (10) shows that if DOMAFFIX (8a) is ranked below FAITH-ΙΟ (8d) and ALIGNMORPH (8b), then it is optimal for the RED tone to be realized within RED, as in the Yoruba gerundives in (4):

98 (10)

Laura J. Downing Yoruba gerundives (4)

RED, Η = mu

FAITH-

IO

ALIGN MORPH

^"a. mi = mu b. mi = mil c. mi = mu

*!

FAITH- B R

DOMAFFIX

*

*

*

*!

*

Constraint ranking optimizing tones aligned within morphemes: FAITH-ΙΟ (8d), ALIGNMORPH (8b) »

FAITH-BR (8C), DOMAFFIX (8a)

Candidate (10a) is optimal, as all input tones are realized in the output, satisfying FAITH-ΙΟ (8d). And each morphological tone is realized within the morpheme that contributes it, satisfying ALIGNMORPH (8b). The nonoptimal candidates (10b) and (10b) both illustrate tonal identity between RED and Base, satisfying FAITH-BR. Candidate (10b) is non optimal, though, as it violates FAITH-ΙΟ (the Base's output mora does have the same tone as in the input). Candidate (10c) is non optimal as it violates ALIGNMORPH (the High tone contributed by the RED is not realized in the output RED string; indeed, it is not realized in the output at all).

3.2.

Base tone melody realized over the reduplicative complex

We have just seen that 'Emergence of the Marked' tonal asymmetries between the RED and the Base can arise when RED realizes a marked tone (or tone melody) which is contributed by the reduplicative construction. 'Emergence of the Marked' tonal asymmetries between the RED and the Base can also arise when RED realizes a marked tone which is contributed by the Base, while the corresponding Base string bears a different tone.

3.2.1. Data An example of this pattern is found in Hausa pluractional verbs. (The pluractional indicates iterative action.) As Newman (2000) shows, one productive way of forming the pluractional is with a CVC- reduplicative prefix. The tone melody assigned to the pluractional construction is based on the

The emergence of the marked

99

verb grade of the Base. The data in (11) illustrate that both the Base and the pluractional form have the tone melody appropriate to the verb grade of the Base. If the Base is two syllables and the tone melody contains three tones, as in (lla-c), the RED pluractional prefix and the corresponding Base syllable have a different tone: (11)

Hausa pluractional verbs (Newman 2000, chapter 55) Base Two syllable Bases (a) bugää (b) säyää (c) käshee (d) jänyee Bases (e) (f) (g)

Pluractional

Gloss

büb-bügää säs-säyää kär-käshee jäj-jänyee

'beat' 'buy' 'kill' 'pull away'

ν 1,HLH v2, LHL v4, HLH v4, HLH

'frighten' 'be afraid' 'gnaw at'

v l , HLH v3, LHL v2, LHL

with more than two syllables tsoorätä tsät-tsöorätä tsoorätä tsät-tsoorätä däägurää däd-däägurää

Grade, melody

Notice that the RED pluractional prefix can have either a High tone or a Low tone, depending on the verb grade melody. Further, the tone of the RED does match the corresponding Base syllable if the Base verb is long enough to realize the verb grade melody. As a result, the tonal asymmetry in ( l l a - c ) cannot be a T E T U effect. Many Bantu languages have a productive process of verb stem reduplication similar to that already illustrated for Yao in (2), above. As we saw, in Yao the Base verb stem tone is realized on one half of the reduplicative complex, and the unmarked tone is realized on the other half, consistent with a TETU analysis. In other Bantu languages, however - like Haya (Hyman & Byarushengo 1984), Kimatuumbi (Odden 1996a), Runyankore (Poletto 1998), Kinande (Mutaka & Hyman 1990) and Shona (Odden 1984) - the Base stem tone (melody) is realized over the entire RED+Base complex. This can result in a marked tone of the Base melody being realized on the RED, while the corresponding Base mora has the unmarked tone. The Ί didn't' verb stem paradigms of Shona shown in (12) illustrate this pattern. Stems in this paradigm contribute a 'grammatical' High tone as well as potentially a lexical High tone to the output tone pattern. As Odden (1984) shows, in toneless verbs, the grammatical High tone is realized on the second syllable of unreduplicated stems (12a,b). Notice that redupli-

100

Laura J. Downing

cated toneless verb complexes of the same length have the same tone pattern: the High tone is realized on the second stem syllble, if the reduplicative complex is considered a single stem. As shown in (12c,d), when the stem has a lexical High tone, the grammatical High tone is realized on the stem-final syllable, and the lexical High tone on the first syllable of the stem. Again, reduplicated High-toned verb complexes of the same length have the same tone pattern, if the reduplicative complex is considered a single stem. (12)

Shona partial verbal reduplication (adapted Odden 1984: 270); only stems are given, with the tone pattern they would have following the INFL string, handakä- Ί didn't...' 9 (a)

Low tone verb stem, 4 syllables -bikisira 'make cook for'

Reduplicated L stem, 4 syllables -bikä-bika 'repeatedly cook'

(b)

Low tone verb stem, 5 syllables -bikisirana 'make cook for each other

Reduplicated L stem, 5 syllables -bikä-bikisa 'make cook repeatedly'

(c)

High tone verb stem, 4 syllables -toreserä 'make take for'

Reduplicated Η stem, 4 syllables -tora-torä 'take frequently'

(d)

High tone verb stem, 5 syllables -toreserana 'make take for each other'

Reduplicated Η stem, 5 syllables -töra-toresä 'make take frequently'

As Downing (2002, 2003) argues, the best explanation for why the stem High tone melody of the Base in languages like Shona is realized over the entire reduplicative complex is that the RED+Base complex is a compound stem. Stem tone assignment treats the compound stem as a single domain of tone realization, just as stress assignment can treat a compound word as a single stress domain in other languages. 10 In both Hausa pluractionals (11) and Shona (12), then, we find tonal asymmetries between the RED and the Base which cannot be explained in terms of TETU. The R E D does not have a fixed tone pattern, and the RED bears a marked tone different from the corresponding Base mora. Instead, the asymmetry is best explained by proposing that the tone melody contributed by the Base takes the entire reduplicative complex as a single domain of tone realization. That is, these tone patterns result when the Base's tone 'dominates' other potential sources of tone realization on R E D (such as

The emergence of the marked

101

tonal identity with the base or TETU) and can be analyzed on analogy with the dominant Affix tone patterns discussed in section 3.1, above.

3.2.2. Formal

analysis

I propose that the 'dominant' Base tone pattern which results when a marked tone contributed by the Base is realized on the RED can be accounted for by properly ranking the constraints in (13), which are nearly identical to the constraints in (8), above: (13)

Constraints defining Base tone assignment to RED+Base (a)

(b)

DOMBASE11 Align(BaseT, Stem/Word): Align the tone melody of the Base within the entire reduplicative complex (Stem for Bantu languages like Shona; Word for Hausa). ALIGNMORPH

Align(MorphT, Morph): Align the tone melody of a morpheme within the morpheme that contributes it to the Input. (c)

FAITH-BR(TONE): For corresponding moras in the Base and RED, each mora of the Base should have the same tone in RED and each mora of RED should have the same tone in the Base.

(d)

FAITH-IO(TONE): For corresponding moras in the Input and Output, each mora of the Input should have the same tone in the Output and each mora of the Output should have the same tone in the Input.

(13a), like D O M A F F I X (8a), optimizes aligning the tone (melody) of the dominant morpheme within the entire reduplicative complex. In the languages with dominant Base tone described in the preceding section, this constraint must be highly ranked, as it defines the dominant tone pattern. A L I G N M O R P H (8b), repeated as (13b), optimizes realizing the tone of each morpheme within the morpheme that contributes it to the input. This constraint must be low ranked in languages where the Base tone is dominant. FAITH-BR (8c), repeated as (13c), is also low-ranked, as the RED does not show tonal identity with the Base when Base tone is dominant. DOMBASE

102

Laura J. Downing

The tableau in (14) exemplifies the analysis of the dominant Base pattern in Shona: (14)

Shona Ί didn't...'(12)

RED=bika, Η

DOMBASE

a. bikä = bikä

*!

b. bika = bikä

*!

FAITH-BR

FAITH-IO

ALIGNMORPH

*

*

* *

^ c . bikä = bika

*

RED=toresa, Η Η d.t0rä=toresä

*!

*

*

*

*

e. tora-toresä ^ f . töra=toresä

*

Constraint ranking optimizing dominant Base tone: DOMBASE » FAITH-BR, FAITH-ΙΟ, ALIGNMORPH In the first set of output candidates, for low-toned stems, candidate (14c) is optimal. The Base High tone is associated with the second mora of the entire reduplicative complex, satisfying the highest-ranked constraint, DOMBASE (13a). Candidate (14a), which illustrates tonal identity, is non optimal as it violates DOMBASE. Candidate (14b), which illustrates TETU, also fatally violates DOMBASE (13a). In both failed candidates, the Base High tone is not associated in such a way that there is a High tone on the second mora of the entire reduplicative complex. In the second candidate set, (14f) is optimal. The Base High tones are associated with the first and final moras of the reduplicative complex, satisfying DOMBASE (13a). This candidate also clearly illustrates the distinction between DOMBASE and FAITH-ΙΟ. Even though both constraints require input tones contributed by the Base to be realized in the Output, FAITH-IO is a correspondence constraint, evaluating similarity between the tone of Base moras in the Input and Output. Candidate (14f) violates FAITH-ΙΟ as the input Base has no associated High tones, while the output does. However, it satisfies DOMBASE, an alignment constraint, as the High tones contributed by the Base are aligned over the entire reduplicative complex. Candidate (14d), which illustrates tonal identity, is non optimal as it violates DOMBASE by copying the Base tone pattern instead of realizing it once, over the entire reduplicative complex. Candidate (14e), which illus-

The emergence of the marked

103

trates TETU, also fatally violates DOMBASE, by realizing the Base tone pattern entirely within the Base stem. The tableau in (15) shows that the same constraints and rankings also account for the Hausa pluractionals: (15) Hausa pluractionals (11) RED=bugaa, HLH a.büb-bügää ^ b . bub-bügää

DOMBASE

FAITH-BR

ALIGNMORPH

*! *

*

Constraint ranking optimizing dominant Base tone: DOMBASE » FAITH-BR, ALIGNMORPH Candidate (15b), in which the Base verb grade tone melody is realized over the entire pluractional form, is optimal, as it satisfies DOMBASE (13a). The competing candidate, which illustrates tonal identity, is non-optimal, as it violates DOMBASE. To sum up this section, there are three important sources for the reduplicative tonal asymmetries which I have labeled the Emergence of the Marked: the RED has a marked tone and the corresponding Base mora an unmarked. The first source is that the RED contributes a marked tone and the Base an unmarked tone to the input, and these tones are realized within the morphemes which contribute them in the output, as in Yoruba gerundives (4). Another is that the RED can contribute a tone melody containing marked and unmarked tones which is realized over the entire reduplicative complex (instead of just within the RED) in such a way that the RED is associated with a marked tone of the melody and the corresponding Base mora with an unmarked, as in Hausa plurals (6). Finally, the Base can contribute a tone melody containing marked and unmarked tones which is realized over the entire reduplicative complex (instead of just within the Base) in such a way that the RED is associated with a marked tone of the melody and the corresponding Base mora with an unmarked, as in Shona (12). Even though the Emergence of the Marked tonal patterns are the opposite of the more familiar TETU patterns, the formal analyses reveal similarities between the two types of grammars. In the Emergence of the Marked grammars, like the TETU grammar (FAITH-IO » MARKEDNESS » FAITH-BR), FAITH-BR must be low ranked to optimize a tonal asymmetry between RED and the Base. However, tonal markedness is ranked below

104

Laura J. Downing

both types of Faithfulness constraints in the Emergence of the Marked grammars, as both the RED and the Base can be associated with a marked tone in the output. What optimizes the Emergence of the Marked is for tone-morpheme alignment constraints ( D O M A F F I X , D O M B A S E , A L I G N M O R P H ) to outrank FAITH-BR. A S we have seen, these constraints can optimize associating a marked tone contributed by either the Base or the reduplicative construction on the RED, potentially leading to a tonal asymmetry with the Base.

4. The autosegmental nature of tone As this brief survey shows, in mismatches between the tone of the Base and the tone of the RED, the RED often has a marked tone while the Base has an unmarked tone. Further, REDs which are arguably affixes (like the Yoruba gerundive (4)) are just as likely to have the marked tone as REDs which are arguably stems. I suggest that two special properties of tone predict this. First, tone is an autosegmental feature. As work like Pulleyblank (1986) shows, an autosegmental property of tone is that it can act as a morpheme independent of segments. This predicts that even an affix without input segmental content like RED can contribute a marked tone melody independent of the Base tone, just like other affixes do. For example, as shown in (16), all the plural affixes in Hausa with fixed segments have a dominant tone melody. It is, then, unsurprising that the reduplicative plural affixes like those illustrated in (6), above, have similar tonal properties: (16)

Hausa fixed segment plurals with dominant plural tone melody (Newman 2000: 431)

(a) (b) (c) (d) (e) (f)

Singular

Plural

Gloss

dali:li hu:la: gun til: küje:rä: bo:kiti kyärkexi:

dali:lai hii:luna: giintattäki: kuje:rii: bo:kitai kyarke:tai

reason cap stub chair bucket wild dog

Secondly, tone is a prosodic feature, regularly associating over a morphologically complex domain, rather than exclusively within the morpheme

The emergence of the marked

105

that contributes it. (See Yip (2002) for a recent overview.) As a result, it is expected that RED can have a marked tone if it is within the domain of realization of that tone, even if the corresponding Base string does not. For example, in presenting the Shona data in (12), above, we noted that the grammatical High tone is realized on particular positions in the stem, depending on the lexical tone of the stem. It is realized on the second stem syllable if the lexical stem tone is Low, but on the final stem syllable if the lexical stem tone is High. As a result, it is not surprising that, if the reduplicative complex is a compound stem, RED will be within the domain of realization of marked tones contributed by the Base that target positions in the stem. These two special properties explain why tonal asymmetries in reduplicative constructions frequently cannot be accounted for by the TETU hypothesis, even though this theory works well for other types of RED-Base asymmetries. Other phonological features do not so regularly define morphemes without segmental support, nor do other features regularly associate to particular positions at some distance from the morpheme or segment which contributes them to the input.

5.

Conclusion

To sum up, the analysis presented here of RED tone realization sheds an interesting light on the interaction of morphology and phonology in defining the output of reduplication. There is a body of recent work - e.g., Downing (1999, 2000, 2002, 2003), McCarthy & Prince (1995, 1999), Urbanczyk (1996, 2000) - arguing that phonological properties of reduplicative morphemes often fall out from their morphological category (Word, Stem, Root, Affix). As we have seen, tonal asymmetries do not conform to the simple prediction that the RED should have less marked phonological material compared to the Base because they are necessarily morphologically and phonologically dependent on the Base. However, REDs do often have tonal properties consistent with other morphemes of the same type. For example, affixal REDs can contribute tones to the input if other affixes do. And if other affixes have dominant tone melodies, as in Hausa, then reduplicative tone melodies can also be dominant. Similarly, in languages like Shona where the stem is the usual domain for tone realization, if the reduplicative complex is a compound stem it can be treated as a single domain for tonal processes that target stems. This finding is echoed in the

106

Laura J. Downing

work of Inkelas (this volume) who argues that the phonological substance of reduplicative forms in many languages is best explained if the reduplicative complex is a compound. Tone realization in reduplicated forms thus reveals the complex morphological and prosodic relation of the RED with its Base. R E D is an independent morpheme with potentially independent tonal properties, but it is dependent on the Base morphologically as it forms a morphological and, potentially, prosodic unit with the Base. As we have seen, both the tonal independence of RED and its morpho-prosodic dependence can be sources of tonal asymmetries which lead to the Emergence of the Marked tone on RED in many African languages.

Notes 1.

2.

3.

I would like to thank Nike Orie, Caro Struijke, Cheryl Zoll and other members of the audience at the Graz Reduplication conference, and an anonymous reviewer for comments which improved the substance and presentation of this paper. Any remaining errors of fact or interpretation are my responsibility. In this paper I am adopting McCarthy & Prince's (1995) proposal that there are two main types of constraints evaluating the well-formedness of output strings: markedness constraints and correspondence (faithfulness) constraints. Markedness constraints penalize marked structure in the output. Correspondence constraints evaluate how faithful (similar) morphologically related strings are to each other. 1-0 correspondence compares an input string with an output string. B-R correspondence constraints compare a RED string with its Base. Morphologically related strings (Inputs and Outputs or Bases and REDs) are often not identical, though. This is explained in Optimality Theory by ranking markedness constraints above the relevant correspondence constraints. The TETU pattern, where unmarked structure is required in the RED although the corresponding Base string contains marked structure, is obtained by the following constraint ranking: FAITHIO » MARKEDNESS » F A I T H B R . See Inkelas & Zoll (2000), Inkelas (this volume) and Pulleyblank (to appear), however, for arguments against B-R correspondence. And see Struijke (2000) for discussion of a third type of correspondence relevant to reduplication: Broad 1-0 correspondence. See Steriade (1995) for discussion of why [+high] vowels are unmarked epenthetic vowels cross-linguistically and Pulleyblank (1986) for discussion of why Mid is the unmarked tone in a three tone system like that of Nupe.

The emergence of the marked 4.

107

Myers & Carleton (1996) develop this proposal for Yao, a language where stem-internal derivational affixes do not bear tone. It is unclear whether their proposal could be extended to account for the Yoruba pattern in (3), as affixes do bear contrastive tone in this language. Also, as Downing (2002, 2003) shows, tonal realization in verbal reduplicative complexes across Bantu languages is best explained if RED is a stem, forming a compound with the Base. See sec. 3.2, below, for further discussion of this point. 5. It is important to note that McCarthy & Prince (1995, 1999) suggest a different conception of the Emergence of the Marked, namely, the grammar defined by reversing the ranking of the Faithfulness constraints that define TETU, to give: Faith B-R » Markedness » Faith I-O. This ranking would optimize marked phonology in the RED which is not found in the language as a whole, and I agree with them in judging this grammar 'pathological'. 6. As one might guess from the number of papers cited in introducing the Yoruba gerundive reduplication pattern, there is much lively debate about most aspects of its analysis except that the High tone of the gerundive is contributed by the prefix. The interested reader should consult the works cited for further discussion. 7. I am following Inkelas's (1998) analysis of non-reduplicative affixal tone in Hausa in using the term "dominant tone" - on analogy with dominant accent to refer to morphological tone (melodies) which replace the tone (melody) of the Base to which the morpheme is affixed. 8. DOMAFFIX (8a) is obviously an abbreviation for a set of constraints which would optimize aligning reduplicative HLHH melody within the entire plural word (instead of just within the reduplicative morpheme which contributes the melody). Details of the Hausa analysis are not developed here, for ease of comparison with reduplicative tone realization in the other languages discussed. 9. In presenting the Shona data in (12), I am abstracting away from a process of tone spread which applies to the leftmost High tones. See Odden (1984), Hewitt & Prince (1989), Downing (1996) and Yip (2002) for further discussion of the surface tone patterns in these forms. 10. As Downing (2002, 2003) shows, the range of cross-Bantu variation in the realization of tone in reduplicated verb stems is best accounted for by defining the RED+Base complex as a compound stem. The interested reader can consult these works for further discussion. 11. DOMBASE (13a) is obviously an abbreviation for a set of constraints which optimize the tone patterns illustrated in the data in (11) and (12). Details of the Shona and Hausa analyses are not developed here, for ease of comparison. See Downing (2003) for a more detailed discussion and formal analysis of Bantu reduplicative tone patterns like that of Shona.

108

Laura J. Downing

References Akinlabi, Akinbiyi 1997 Patterns of tonal transfer I. Paper presented at ACAL 28, Cornell University, July 12, 1997. 2000 Asymmetries in reduplicative and nonreduplicative defaults. Paper presented at ACAL 31, Boston University. Akinlabi, Akinbiyi and Eno Urua 2000 Tone in Ibibio verbal reduplication. In Proceedings of the 2nd World Congress of African Linguistics Leipzig 1997, H. Ekkehard Wolff and Orin D. Gensler (eds.), 279-291. Köln: Rüdiger Köppe. Alderete, John, Jill Beckman, Laura Benua, Amalia Gnanadesikan, John McCarthy and Suzanne Urbanczyk 1999 Reduplication with fixed segmentism. Linguistic Inquiry 30: 327364. Awoyale, Yiwola 1989 Reduplication and the status of ideophones in Yoruba. Journal of West African Languages 19: 15-34. Beckman, Jill N. 1998 Positional Faithfulness. Ph.D. dissertation, University of Massachusetts-Amherst. Capo, Hounkpati B.C. 1991 A Comparative Phonology of Gbe. Berlin/New York: Foris. Clark, Mary M. 1990 The Tonal System of Igbo. Dordrecht: Foris. Cohen, Kevin Bretonnel. 2000 Aspects of the Grammar ofKukii. Munich: LINCOM EUROPA. Downing, Laura J. 1996 The Tonal Phonology ofJita. Munich: LINCOM EUROPA. 1999 Morphological constraints on Bantu reduplication. Linguistic Analysis 29 (1-2): 6-46. 2000 Morphological and prosodic constraints on Kinande verbal reduplication. Phonology 17: 1-38. 2002 Tone (non-) transfer in Bantu verbal reduplication. Proceedings of the Workshop on the Typology of African Prosodic Systems. Bielefeld Occasional Papers in Typology 1: 41-54. 2003 Compounding and tonal non-transfer in Bantu languages. Phonology 20(1): 1-42. Hewitt, Mark and Alan Prince 1989 OCP, locality, and linking: the N. Karanga verb. WCCFL 8: 176191. Hyman, Larry M. and Ernest Rugwa Byarushengo 1984 A model of Haya tonology. In Autosegmental Studies in Bantu Tone, G.N. Clements and J. Goldsmith (eds.), 53-103. Dordrecht: Foris.

The emergence of the marked

109

Hyman, Larry M. and Al Mtenje 1999 Prosodic Morphology and tone: the case of Chichewa. In The Prosody-Morphology Interface, Rene Kager, Harry van der Hulst and Wim Zonneveld (eds.), 90-133. Cambridge: Cambridge University Press. Inkelas, Sharon 1998 The theoretical status of morphologically conditioned phonology: a case study of dominance effects. In Yearbook of Morphology 1997, Geert Booij and Jaap van Marie (eds.), 121-155. Dordrecht: Kluwer Academic Publishers. this vol. Morphological Doubling Theory: Evidence for morphological doubling in reduplication. Inkelas, Sharon and Cheryl Zoll 2000 Reduplication as morphological doubling, ms., UC-Berkeley and MIT. Levergood, Barbara 1991 The role of the mora in Maasai. ms. University of California-San Diego. McCarthy, John J. and Alan S. Prince 1993 Prosodic Morphology I: constraint interaction and satisfaction, ms. University of Massachusetts, Amherst, MA and Rutgers University, New Jersey. 1995 Faithfulness and reduplicative identity. UMOP 18: Papers in Optimality Theory. 249-384. 1999 Faithfulness and identity in Prosodic Morphology. In The ProsodyMorphology Interface, Rene Kager, Harry van der Hulst, and Wim Zonneveld (eds.), 218-309. Cambridge: Cambridge University Press. Mtenje, Al 2002 An Optimality Theoretic account of CiYao verbal reduplication. To appear in Proceedings of ACAL 33. Mutaka, Ngessimo and Larry M. Hyman 1990 Syllables and morpheme integrity in Kinande reduplication. Phonology 7: 73-119. Myers, Scott and Troi Carleton 1996 Tonal transfer in Chichewa. Phonology 13: 39-72. Newman, Paul 2000 The Hausa Language: An Encyclopedic Reference Grammar. New Haven: Yale University Press. Niepokuj, Mary K. 1991 The Historical Development of Reduplication, with Special Reference to Indo-European. Ph.D. dissertation, University of CaliforniaBerkeley.

110

Laura J. Downing

Odden, David 1984 Stem tone assignment in Shona. In Autosegmental Studies in Bantu Tone, G.N. Clements and J. Goldsmith (eds.), 255-280. Dordrecht: Foris. 1996a The Phonology and Morphology of Kimatuumbi. Oxford: Clarendon Press. 1996b Patterns of reduplication in Kikerewe. OSU WPL 48, 111-148. Orie, Olanike-Ola 1997 Benue-Congo Prosodic Phonology and Morphology in Optimality Theory. Munich: LINCOM EUROPA. Poletto, Robert E. 1998 Topics in Runyankore Phonology. PhD dissertation, The Ohio State University. Pulleyblank, Douglas 1986 Tone in Lexical Phonology. Dordrecht: Reidel. 1988 Vocalic underspeciflcation in Yoruba. Linguistic Inquiry 19: 2 3 3 270. forthc. Patterns of reduplication in Yoruba. In The Nature of the Word: Essays in Honor of Paul Kiparsky, Sharon Inkelas and Kristin Hanson (eds.), Cambridge, Mass.: MIT Press. Smith, N.V. 1969 The Nupe verb. African Language Studies X: 90-160. Steriade, Donca 1988 Reduplication and syllable transfer in Sanskrit and elsewhere. Phonology 5: 73-156. 1995 Underspeciflcation and markedness. In The Handbook of Phonological Theory, John Goldsmith (ed.), 114-174. Oxford: Blackwell. Struijke, Caro 2000 Existential Faithfulness: A study of reduplicative TETU, feature movement, and dissimilation. Revision of Ph.D. dissertation, University of Maryland. Urbanczyk, Suzanne 1996 Patterns of Reduplication in Lushootseed. Ph.D. dissertation, University of Massachusetts-Amherst. 2000 Reduplicative form and the Root-Affix asymmetry, to appear: NLLT. Wilbur, Ronnie 1973 The phonology of reduplication. PhD dissertation, University of Illinois at Urbana-Champaign. [distributed by the IULC], Yip, Moira 2002 Tone. Cambridge: Cambridge University Press.

Reduplication and consonant mutation in the Northern Atlantic languages* Fiona Mc Laughlin

1.

Introduction

Recent constraint based approaches to reduplication have focused variously on the phenomenon as either a predominantly phonological problem, as in McCarthy & Prince (1995) and subsequent literature within the framework of Correspondence Theory, or as a predominantly morphological problem, as in Inkelas & Zoll's Morphological Doubling Theory (2000; this volume) and related approaches such as Hyman et al. (1998). One of the principal tenets of Correspondence Theory is that a relationship of phonological correspondence or identity exists between a base (B) and a reduplicant (R), requiring them to be identical. BR correspondence accounts for unexpected phonological identity between the two insofar as it provides an explanation for problematic cases of so-called underapplication, overapplication, and especially "backcopying" whereby a base is altered to more closely resemble the reduplicant. Morphological Doubling Theory, on the other hand, claims only morphosemantic identity between the two elements in a reduplicative construction, each of which is constrained independently by a distinct cophonology which accounts, among other things, for the discrepancy in shape between the base and the reduplicant in partial reduplication. Phonological identity between base and reduplicant may, and often does, occur as a side effect of morphosemantic identity, but Morphological Doubling Theory assumes no phonological correspondence between them. Empirical evidence for either of these two theories of reduplication must come from the interaction of reduplication with other phonological phenomena. This chapter offers a unique case study of the interaction of reduplication with consonant mutation in the Northern Atlantic languages, and addresses its implications for a theory of reduplication. Pulaar (a dialect of Fula) and Seereer-Siin, two Northern Atlantic (Niger-Congo) languages, are characterized by morphologically conditioned

112

Fiona Mc Laughlin

stem-initial consonant mutation which interacts with reduplication in various ways. The approach to consonant mutation adopted here is that it involves the affixation of a phonological feature to the root node of a stem consonant. The reduplicant in both Pulaar and Seereer-Siin is a prefix, so the initial consonant of the reduplicative construction is in exactly the right morphological environment to undergo consonant mutation, as shown in the examples in (1). While Pulaar shows well-behaved or expected interaction between consonant mutation and reduplication, in one such pattern of featural affixation in Seereer-Siin the floating feature surfaces optionally on the base in a case of apparent featural transfer or backcopying. This is illustrated in (2), where the forms otiiriw and opeefec are the expected ones and the righthand forms otiitiw and opeepec are unexpected in that they exhibit backcopying. 0)

(2)

Pulaar

Seereer-Siin

hul-de to fear

kul-hul-i frightening things

seer-de to separate

ceer-seer-o divorcee

riw to weave

otii-riw ~ otii-tiw weaver

fee to dance

opee-fec ~ opee-pec dancer

In a correspondence theoretic analysis, transfer effects such as those seen in the Seereer-Siin examples in (2) can be explained by BR correspondence in which the base is modified to maintain maximum identity with the reduplicant. True backcopying, however, cannot be accommodated by Morphological Doubling Theory since there is no relationship of phonological correspondence between base and reduplicant. In this paper I describe and analyse the interaction of consonant mutation and reduplication in Pulaar and Seereer-Siin and discuss the implications of the case of backcopying illustrated in (2) for theories of reduplication. As I will show, these instances of backcopying raise substantial problems for BR correspondence, but for morphological reasons they also resist the type of infixation analysis proposed by Inkelas and Zoll (2000) for cases of apparent backcopying in Chumash and Tagalog. The backcopying problem is compounded by the variation found in Seereer-Siin reduplication, which also demands a solution. I propose that the facts can

Reduplication and consonant mutation

113

be accounted for in a straightforward way by appealing to the interaction of alignment and licensing constraints in a manner similar to that proposed by Piggott (1997; 2000) for other types of featural affixation. But crucially, my analysis can only succeed if reduplication is viewed as stem juxtaposition (Hyman et al. 1998) rather than the derivation of a reduplicant from a base. As I will show, the behavior of the floating feature in Seereer-Siin turns out to be a simple consequence of the fact that by virtue of multiple association to a root node, a projected featural affix can surface on more than one segment without being duplicated. The same is not true of segmental affixes. Not only does this account solve the problems inherent in other approaches, the same licensing and alignment constraints play a role elsewhere in the grammar, thus the need for a reduplication-specific mechanism is eliminated. The organization of the paper is as follows: In §2 I present consonant mutation as featural affixation to a root node, describe patterns of mutation in Pulaar and Seereer-Siin within this purview, and discuss the constraints that govern featural affixation. In §3 I discuss the interaction of consonant mutation with reduplication in each of the two languages; I then make the case in §4 for an analysis that rejects both BR-correspondence and infixation, and favors the interaction of alignment and licensing constraints in conjunction with stem juxtaposition.

2.

Consonant mutation

Stem-initial consonant mutation occurs in both Pulaar and Seereer-Siin in contexts of verb-to-noun and noun-to-noun derivation, and in nominal and verbal inflection. In verbs, mutation is conditioned primarily by number and focus, while in nouns it is conditioned by noun class. The discussion of mutation in this section will focus on the context of nominal inflection since that is where the fullest range of mutations occurs in both languages. There are twenty-one noun classes in Pulaar and sixteen in Seereer-Siin. Of the sixteen classes in Seereer-Siin, eleven have overt class prefixes of the shape (C)V in addition to conditioning consonant mutation. These are illustrated in (3), where the stem-initial consonant alternates between a voiceless stop, [k], and a voiced prenasalized stop, [ng], (3)

Seereer-Siin 'man'

okoor fongoor

class 1 class 13

(human singular) (diminutive plural)

114

Fiona Mc Laughlin

Morpheme boundaries between prefix and stem do not occur between segments, but rather between features at a subsegmental level. The morphological breakdown of the forms in (3) are as follows: (4)

okoor fongoor

/o [-continuant]/ i n f l + /koor/ N /fo [nasal]/ ^ f l + /koor/ N

Affixes like these which consist of one or more segments as well as a featural exponent are also attested in the literature for Terena, an Arawakan language (Bendor-Samuel 1960; Harden 1946; Piggott 1997, 2000) and Chaha, a Gurage (Afroasiatic) language (Johnson 1975; McCarthy 1983; Rose 1993). Like Pulaar, the remaining five Seereer-Siin classes have no overt prefixes, but they nevertheless condition consonant mutation. In contradistinction to Seereer, Pulaar has obligatory class suffixes. Examples of inflected nouns with no overt class prefix are given in (5) for both languages. (5)

Seereer-Siin 'man' goor Pulaar 'man'

gorko wor\e ngoron

class 2

(human plural)

class 1 class 2 class 21

(human singular) (human plural) (diminutive plural)

The morphological breakdown of an inflected noun such as the Pulaar class 1 form for 'man' is as follows: (6)

gorko

/[-continuant]/ INFL + /wor/ N + /ko/ INFL

Although recent approaches to consonant mutation (Ni Chiosäin 1991; Elzinga 1996, Gnanadesikan 1997, Kibre 1997) have yielded no consensus on how best to treat the phenomenon, many autosegmental approaches to consonant mutation in Fula have viewed it as a type of featural affixation in which a floating feature attaches to the root node of a consonant (Lieber 1984, 1987; Paradis 1987). I adopt this approach to consonant mutation not only because of the ease of analysis it lends to a complex phenomenon, but also because it is compatible with the comparative and historical facts. Historical evidence (Greenberg 1977) strongly suggests that in the Atlantic languages, at least, consonant mutation is the residue of once overt prefixa-

Reduplication and consonant mutation

115

tion. As illustrated in (4), Seereer-Siin retains some of these overt prefixes in conjunction with consonant mutation, but the other Northern Atlantic languages do not. Fula, including the Pulaar dialect, retains consonant mutation but in conjunction with class suffixes, while Wolof retains only traces of consonant mutation in inflectional contexts (Mc Laughlin 1997).

2.1. Consonant mutation in Pulaar As is familiar from the abundance of literature on the topic (Anderson 1976, Arnott 1970, Churma 1986, Klingenheben 1927, Lieber 1984, 1987, Paradis 1987, Skousen 1972, inter alia), the initial consonant of nominal stems in Pulaar and many other Fula dialects may alternate between three homorganic variants known conventionally as grades, after Arnott (1970). In a full mutation set the grades consist of a continuant, a plain stop, and a prenasalized stop. There are nine mutation sets in Pulaar, given in (7), and any given noun class conditions the same grade, either a, b, or c, regardless of the mutation set. (7)

Pulaar gradation sets a. b. c.

cont stop nasal

LABIAL w f b Ρ mb Ρ

CORONAL r s y d c j nd c nj

DORSAL h y k g k ng

w

?

g ng

g ng

Voicing is constant over a gradation set so that the entire set is either voiced or voiceless. Consequently, plain voiceless stops appear in the c-grades of voiceless gradation sets because only voiced stops can be prenasalized in Pulaar. Simple nasal stops, the liquid [1], and the implosive consonants in Pulaar do not participate in the mutation system. Examples of nominal paradigms that exhibit three homorganic variants are given in (8). (8) Class 1: human singular

'woman' debbo

Class 2: human plural Class 21: augmentative singular

rewße ndewon

'man' gorko worße ngoron

116

Fiona Mc LaughI in

Almost all studies of Pulaar or other Fula dialects (Anderson 1976; Churma 1986; Paradis 1987; and Niang 1997), with the exception of Skousen (1972) and Elzinga (1996), agree that continuant-initial forms are basic in noun stems. 1 Mutation thus results from the affixation of either a [continuant] prefix to a continuant, thereby changing it to a stop, or a [nasal] prefix, changing it to a prenasalized stop. There are no affixes that condition the a-grade. Noun classes that "condition" a continuant-initial stem are simply characterized by a null class prefix. That continuants are basic is based on systematic patterns of stem behavior, including those of nonmutating stems that we would otherwise expect to undergo mutation, as well as corpus-external evidence from French loanwords (Mc Laughlin 2000b). 2 As we will see in §3.1, the fact that uninflected noun stems are continuant initial will bear on the discussion of Pulaar reduplication.

2.2. Consonant mutation in Seereer-Siin There are two types of consonant mutation in Seereer-Siin (Mc Laughlin 1994, 2000a). One is similar to the Pulaar pattern in which the homorganic variants consist of a continuant, a stop, and a prenasalized stop. This pattern, given in (9), is known as continuancy mutation. In contradistinction to Pulaar, however, voicing is not necessarily constant across a gradation set. All consonants are prenasalized in the c-grade, and all are voiced in that grade as well. The second type, known as voicing mutation, involves an alternation between homorganic plain voiced stops, plain voiceless stops and prenasalized stops, and is given in (10). Seereer-Siin has a relatively unusual set of voiceless implosives 3 that participate in the mutation system, although they do not undergo prenasalization due to constraints on featural coocurrence. As mentioned in §2.1, Pulaar implosives do not participate in the mutation system. (9)

Continuancy mutations

a. b. c.

continuant stop nasal

LABIAL

CORONAL

DORSAL

w b mb

r t nd

w k ng

f ρ mb

s c nj

X q nG

h k ng

Reduplication and consonant mutation (10)

117

Voicing mutations

a. b. c.

voiced voiceless nasal

LABIAL

CORONAL

DORSAL

b ρ mb

d t nd

j c nj

6 β β

cf f f

ί c cf

k ng

Any given stem will undergo only one type of consonant mutation. Examples of nominal paradigms in Seereer-Siin cognate with the Pulaar examples in (8), are given in (11). Note that the stem for 'woman' undergoes continuancy mutation while that for 'man' undergoes voicing mutation. (11)

'woman' otew rew ondew

'man' okoor goor ongoor

Class 1: human singular Class 2: human plural Classl2: augmentative singular

Based on the patterns of stem behavior analyzed elsewhere (Mc Laughlin 1994), stems that undergo continuancy mutation, like the stem for 'woman' in (11), are continuant initial in their uninflected form, while those that undergo voicing mutation, like the stem for 'man' in (11), are voiceless stop initial in their basic form. The floating features involved in class prefixation, then, are [+voice] for those that condition the a-grade, [-continuant] for those that condition the b-grade, and [nasal] for those that condition the c-grade. Like Pulaar, Seereer-Siin also has non-mutating and partially mutating stems. These facts are relevant to the analysis of reduplication in §3.2. Table 1 sums up the relevant characteristics of Seereer-Siin and Pulaar inflection as they relate to consonant mutation.

118

Fiona Mc Laughlin

Table 1. Morphological and phonological characteristics of noun class inflection in Pulaar and Seereer-Siin

Overt class prefixes Mutation conditioned by noun class Class suffixes Affix conditioning a-grade Affix conditioning b-grade Affix conditioning c-grade Grade of uninfected stem(s)

Pulaar

Seereer-Siin

no

eleven of sixteen classes

yes yes none [-continuant] [nasal] a, continuant

yes no [+voice] [-continuant] [nasal] a, continuant (continuancy mutation) b, voiceless stop (voicing mutation)

2.3. Featural affixation and licensing The affixation of a featural morpheme or a segmental morpheme with featural exponence such as those described in the preceeding sections has been treated in broadly similar but distinct ways within a constraint based framework (Akinlabi 1996; Piggott 1997, 2000; Zoll 1998). These approaches involve two basic criteria, namely licensing and alignment. While Akinlabi (1996) advocates a separate feature-specific alignment constraint, Piggott (2000) rejects his proposal as unnecessary, favoring instead an analysis that allows the interaction between a more general alignment constraint and licensing constraints to account for problematic data involving mobile featural morphemes such as the [labial] and [high] morphemes found in Chaha.

2.3.1. Licensing and alignment in Pulaar In Pulaar, class is marked by both a featural prefix and an overt segmental suffix. The discussion here will be limited to the class prefix, although the class suffixes are no less interesting in their behavior. The class prefix consists of a floating feature that surfaces attached to a root node in stem-initial

Reduplication and consonant mutation

119

position. Such floating features or subsegments, defined as an undominated F-node (Archangeli & Pulleyblank 1994) by Zoll (1998) must be parsed. This is accomplished by the M A X constraint in (12). (12)

MAX(subseg) (Zoll 1998) Every subsegment in the input has a correspondent in the output.

In accordance with the principles of prosodic licensing a feature must be licensed by a larger prosodic unit, in this case the root node that defines segments. The constraint that licenses featural affixation is given in (13). (13)

FEAT-LIC A feature must be licensed by a root node

Not all segments, however, are legitimate licensors for all features. For example, in Pulaar (and Seereer-Siin) voiceless consonants cannot license the feature [nasal]. Such restrictions are handled by independent constraints on featural cooccurence such as that in (14) which rules out voiceless prenasalized stops. (14)

NAS/VCE If [nasal], then not [-voice]

The relevant alignment constraint in Pulaar, given in (15), ensures that the class prefix be aligned with the left edge of the grammatical word. (15)

ALIGNClass-L, GrWord-L The left edge of the class marker is aligned with the left edge of the grammatical word The class marker is a prefix

Featural affixation necessarily incurs a violation of a featural identity constraint that maps between input and output, since the segment in the output will differ from that in the input. The umbrella constraint in (16) will stand in for all featural affixes. (16)

IDENT-IOf Correspondent 10 segments are identical for the feature f

120

Fiona Mc Laughlin

Relevant constraints are ranked as follows in Pulaar: ALIGN-L, FEAT-LIC « MAX-subseg « IDENT-IOf. An illustration of their effects is given in the tableau in (17) for the class 1 form of the word for man, gorko. The suffix is ignored. In candidate (b) the floating feature is still floating, and therefore unlicensed, while in candidate (c) it is simply unparsed. The affixation of the feature in candidate (a) results in mutation to a stop and is thus the winning candidate. (17) [-cont], wor ^ f a. gorb. [-cont] worc. wor-

ALIGN-L

FEAT-LIC

MAX-subseg

IDENT-IOf *

*!

*!

2.3.2. Licensing and alignment in Seereer-Siin The constraints given in §2.3.1 for Pulaar are equally relevant for consonant mutation in Seereer-Siin. Seereer, however, exhibits the influence of an additional alignment constraint, given in (18), that aligns the right edge of the class affix with the left edge of the stem. (18)

ALIGNStem-L, Class-R Align the left edge of the stem with the right edge of the affix

As we shall see, this morpheme specific constraint in Seereer-Siin plays a crucial role in explaining backcopying in Seereer. The absence (or extremely low ranking) of a similar constraint in Pulaar militates against backcopying in that language.

3.

Reduplication

Reduplication in both Pulaar and Seereer-Siin is prefixal. This places the initial consonant of the reduplicant in the right position to undergo consonant mutation as a consequence of inflection, resulting in the interaction of consonant mutation with reduplication. Coincidentally, Seereer-Siin provides a counterexample to the tendency documented by Hyman et al.

Reduplication and consonant mutation

121

(1998: 35) that the reduplicant will tend to be suffixed to a base when the base has a prefixing structure, and vice versa. Most notably, they cite a series of Bantu languages that follow this tendency, including Kinande where verb reduplication is prefixal but noun reduplication is suffixal. Kinande verbs are characterized by suffixed verbal extensions, thus the reduplicant is prefixed; but nouns take prefixal class markers, thus the nominal reduplicant is suffixed. In Seereer-Siin, although there are prefixed class markers, the reduplicant is nonetheless also prefixed.

3.1. Reduplication in Pulaar Reduplication is generally a limited and unproductive process in Pulaar. Some idiomatic lexical items that involve noun-to-noun or verb-to-noun derivation have reduplicative forms that are derived from a transparent base, such as the examples in (19). While verbs also undergo consonant mutation, conditioned primarily by number and focus, the forms here are given in their continuant-initial infinitival form, which is the same as the underlying form. ngar-7ar-di baal-waal-o kul-hul-i ceer-seer-o kal-hal-di kin-hin-ol ceecf-see(f-lu

'volunteer plant' 'person from Waalo' 'frightening things' 'divorcee' 'male animal' 'thonged sandal' 'signs of approach of dry season'

lar-de waalo hul-de seer-de kal-le hin-ere ceed-u

'to come' 'Waalo region' 'to fear' 'to separate' 'testicles' 'nose' 'dry season'

Like reduplicative forms in certain Bantu languages (Mutaka & Hyman 1990; Downing 2000), the shape of the reduplicant in these examples is determined by morphological considerations. The entire stem must be reduplicated regardless of its shape, CVC or CVVC, resulting in a reduplicant whose prosodic shape is not fixed and does not fit a set CV template. A more consistent and productive pattern of reduplication can, however, be found in Pulaar patronymics. Patronymic reduplication derives a nominal form meaning an individual or individuals of a certain last name, eg: a Kennedy, from that last name. The resulting forms are inflected for noun class 1, human singular, or noun class 2, human plural. Class 1 conditions a stop, while class 2 conditions a continuant.

122

Fiona Mc Laughlin

(20) 'Sail' 'Sy' 'Sow' 'Watt' 'Wane' 'Hathie' 'Agne'

patronym sal sih soh wat wan lac Ίαή

class 1 calsalo ciisiho coosoho batwato banwano gac ?aco gan?ano

class 2 salsalße siisii ßc~ soosooße watwatße warm an ße ?ac?aße Pan Pan ße

Two constraints govern the shape of the reduplicant: as in the reduplicative forms in (19) the entire stem - in this case the patronym - must be reduplicated, but in contradistinction to the forms in (19) the reduplicant must also be bimoraic. Consequently, patronyms that exceed two moras are not reduplicated, as the examples in (21) show. (21) 'Pam' 'Ndiathie' 'Wone'

patronym paam njaac woon

class 1 paamo njaaco goono

class 2 faamße njaac 6e woonße

The pattern of consonant mutation in the examples in (20) is the expected one. The inflectional class marker, [-continuant], is prefixed to the reduplicated form, thus the stem-initial consonant becomes a stop in class 1 forms, but remains unaffected in class 2 forms that have a null class prefix. This is well-behaved reduplication in that there are no overapplication effects and no transfer of the feature, [-continuant], from the reduplicant to the base. All examples in (20) are derived from continuant-initial patronyms, but what happens when stop-initial patronyms such as those in (22) are are reduplicated? (22) 'Kane' 'Ka' 'Dia' 'Bä'

patronym kan kah jah bah

class 1 kanhano kaakaho jaajaho baabaho

class 2 hanhanße haahaaße jaajaaße baabaaße

These forms, too, can be explained in a straightforward way. The first two examples in (22), both derived from [k]-initial patronyms, reveal a crucial aspect of Pulaar morphology. The reduplicative forms are the result of

Reduplication and consonant mutation

123

nominal derivation which, if they are mutating stems, requires them to be continuant-initial. 6 Thus, in the noun-to-noun derivation from [kan] to [kanhano] we get a derived reduplicative stem:[hanhanj which is then inflected for class 1, yielding [kanhano]. It is crucial to recognize that the [h] surfaces in the last two forms in (23) as the result of nominal derivation and constraints on nominal stems rather than as the result of affixation from noun class inflection. This now helps explain the third and fourth examples. These are non-mutating stems, of which there are many in Pulaar as abundantly attested in the literature. A possible reason for their failure to undergo mutation is that both the voiced palatal stop [j] and the voiced bilabial stop [b] are mutations of consonants that have two possible stop forms: namely the glides [y] and [w] respectively, [y] becomes either [j] or [g], while [w] becomes either [b] or [g]. Creating a glide (continuant) initial nominal stem would introduce an element of unpredictability into the paradigm.

3.2. Reduplication in Seereer-Siin Reduplication in Seereer-Siin is used to derive both an agent noun from a verb and the nominal form meaning 'native of a place' from a place name, as illustrated in (23).

'write' 'work' 'stutter'

Fatick Diakhao Ngues

verb

class 1

class 2

bind jal ga?

opiibind ocaajal okaaga?

biibind jaajal gaaga?

place

class 1

class 2

fatik jaxaaw nGees

opaafatik ocaajaxaaw oqeexees1

faafatik jaajaxaaw xeexees

'writer' 'worker' 'stutterer'

'one from Fatick' 'one from Diakhao' 'one from Ngues'

The reduplicant in both cases is a bimoraic syllable of the form CVV where VV is a long vowel. For stems containing short vowels, that vowel is lengthened in the the reduplicant. The discussion of the interaction of reduplication with consonant mutation will focus on the agent nouns since a full range of relevant data is not available for place name derivatives. As the reduplicated forms in (24) and

124

Fiona Mc Laughlin

(25) illustrate, the patterns differ according to the type of mutation the stem undergoes. In stems that undergo continuancy mutation, like those in (24), there is free variation between forms in which featural transfer takes place from reduplicant to base and those in which it does not, while in stems that undergo voicing mutation, like the examples in (25), there is no featural transfer from reduplicant to base.

fl? reef riw xaf xoox

'search' 'kill' 'dance' 'act' follow' 'weave' 'shoot' 'cultivate'

obaa-waacf obaa-war opee-fec opii-fi? otee-reef otii-riw oqaa-xaf oqoo-xoox

~obaa-baa fou 'to crawl') to modify the verb ute 'to hit' (< utu). If fou-fou still remained a lexical verb, the clause would mean 'the messenger brandished the umbrella while crawling (on the ground)', which does not make any sense in this context. Adverbialized reduplication can take neither any linking particles nor direct objects, which implies the concomitant development of adverbial and verbal functions of fafu 'to crawl' (i.e. 'layering'). As examined later in detail, phonological change accompanied by the change from verbal to adverbial can be considered as evidence of grammaticalization. As a result, in order to distinguish between adverbial and clauselinking functions, I will refer to these semantic and morpho-syntactic characteristics of verbal reduplication. Of course, there may be unclear examples at any synchronic stage, which is to be discussed later as being on the cline of grammaticalization.

2.2. V (LF) + V (LF) The other major type is V (LF) + V (LF) which also has adverbial and clause-linking functions. However, this type is different from V (SF) + V (SF) in the distribution of possible morpho-syntactic forms. For example, the particles -ni and -to can only be attached to the adverbial function of V (LF) + V (LF), as in (3) and (4), and the linking particle -te tells us that the form is used in the clause-combining function, as in (5). (3)

kono koto-domo DEM.this thing-PL mo ara-zu PT COP-NEG tukuri make

tada just

kano DEM.that

tiri-diri-ni disperse(LF)-disperse(LF)-PT tukusi name.of.a.place

ni-te LOC-PT

atume-sase-tamafi-keru collect-HON-HON-PST

'These (poems and songs) were not just (taken over) separately, (but the minister) wrote and collected (all the poems) in Tukusi.' (12C Ookagami: 96)

288

Reijirou Shibasaki

(4)

tune always

sira-nu know-NEG

kure(LF)-knre(LF)-io get.dark-get.dark-PT

(5)

miti road

no GEN

nagate long.road

ikani how

ka PT

ika-mu go-will

wo ACC

'How can (I) go (along) a long road in a gloomy mood?' (8C Man 'yoo: 888) mafagi ga tuyu wo wake-wake-te... a.certain.weed NOM drops ACC divide(LK)-divide(LK)-and aki autumn

to COMP

nomi only

kos ο EMPH

sir-are-kere know-POTEN-PERF

'The weed shed drops, and (we) can know that (it is) just autumn.' (15/16C Otogi, Urasima: 165) In (3), the reduplicated verb tiri-diri 'disperse-disperse' accompanied by the particle -ni creates a state or manner 'separately', and the whole phrase modifies the verb ara-zu 'not to be' as an adverbial. In (4), the verb kuru 'to get dark' comes to render 'in a gloomy mood' through reduplication. The reduplicated verb wake-wake in (5) is an example of linking function. It carries its direct object tuyu 'drops' as an argument and depends on the following independent verb sir-are-kere for the entire tense and aspect. 3 The form V (LF) + V (LF) can be suffixed by auxiliaries. In (6), the reduplicated verb miti-miti 'abound-abound' is suffixed with the perfective auxiliary -taru and the whole phrase miti-miti-taru (SF) 'have gathered all over' functions as a final verb. (6)

tuwamono-domo warrior-PL

fima space

fazama space

mo PT

nau not.to.be

zo PT

miti-miti-taru abound(LF)-abound(LF)-PERF 'Warriors have gathered all over (there, so) there is hardly a place to step.' (\?>C Heike: 124) Since auxiliaries also have linking forms, it is grammatically possible to construct an example where the V (LF) + V (LF) + aux form takes a clauselinking function. However, it seems to be quite rare, because the linking form of the verb often reduces the range of tense-aspect possibilities in comparison to verbs in final clauses from a cross-linguistic perspective

On the grammaticalization of verbal reduplication in Japanese

289

(Longacre 1985: 263). Tense-aspect is expressed by the auxiliary in Japanese; therefore, it is reasonable that the verb form with any auxiliary disfavors being used to combine clauses. Such a construction is not found among my examples. The formation V (LF) + V (LF) sometimes occurs with neither particles nor auxiliaries attached, as in Table 1. Since morpho-syntactic cues cannot differentiate whether a given form is an adverbial or a clause-linking function, it is necessary to examine two points in the same way as in V (SF) + V (SF), i.e. whether the form carries any semantically required arguments of a given reduplicative verb and whether the form undergoes any semantic change into an adverb. Consider the following examples. (7)

simogare frozen

no kusamura PT weeds

ni naki tukusi-taru LOC chirp finish-NM

no GEN

kofe voice

tafe-dafe cease(LF)-cease(LF)

musi insect

kikofe-te hear-and

'In the frozen weeds, (certain) unseasonable chirping can be heard discontinuously and...' (14C Tofazugatari, vol.5: 213)

(8)

yosinakaru-beki inconvenient-INFER

waza situation

zo PT

kasi FP

nado COMP

omofi-omofi think(LF)-think(LF)

fazama chink

yori nozoke-ba from peek(REAL)-when

'While thinking 'What an inconvenient situation that would (be)', (the man) peeked from the chink (in the hide)...' (11C Tutumichuunagon: 479-80) In (7), the reduplicated verb tafe-dafe 'cease-cease' yields an adverbial meaning 'discontinuously'. Suppose that the verb tafe-dafe still remains a lexical verb, the whole expression becomes meaningless: 'chirping ceased and could be heard.' Thence, the verb tafu 'to cease' changes its aspectual property from perfective into imperfective through reduplication. In (8), the reduplicated verb omofi-omofi takes the complement clause yosinakarubeki waza zo kasi nado as an argument. Otherwise, this complement clause is of no effect in this context.

290

Reijirou Shibasaki

2.3. V (LF) + V (SF) and V (LF) + V (IR) + ba The third and fourth types are quite rare (two examples for the former and one example for the latter in my texts), and the following are the examples. (9)

kunifara spacious.plain

wa TOP

keburi smoke

tati-tatu stand(LF)-stand(SF)

unafara spacious.sea

wa TOP

kamame seagull

tati-tatu stand(LF)-stand(SF)

'Smoke is going up (lit. smoke stands and stands) from the spacious plains, (and) seagulls are flying up (lit. stand and stand) from the spacious sea.' (8C Man 'yoo: 2) (10)

uwe-si-uwe-ba plant(LK)-PT-plant(IR)-if

aki autumn

naki is.not

toki time

ya PT

saka-za-ramu bloom-NEG-INFER 'If (I) plant (a chrysanthemum) firmly, (it) would not bloom until autumn.' (10C Zse: 176) In (9), two finite clauses are juxtaposed, and the verbal reduplication tatitatu 'stand-stand' is used as a finite verb in both clauses, indicating 'to go up again and again' (iterative) or 'to go up here and there' (distributive). The first form tati is a linking form, while the second form tatu is a sentence-final form. As a whole, this verbal reduplication can function as a finite verb (cf. Hachiya 1998: 63). In (10), the particle -ba is connected to the second part of the reduplicated verb that takes an irrealis form and the whole clause indicates a conditional. The emphatic particle si, which is inserted between the two verbs, seems to be grammatically optional (or a fixed phrase) because it is used just to retain a rhythm in this example. 4

On the grammaticalization of verbal reduplication in Japanese

291

2.4. Other forms of V ( L F ) + V ( L F ) There also exist more complex formations of verbal reduplication of this type, as illustrated in Table 2. Table 2. Types of verbal reduplication 2 Form

Function

Example

VI (LF) V2(LF) + VI (LF) V2(LF)

ADV

osi-kafesi-osi-kafesi 'repeatedly'

VI(LF) V2(LF) + VI(LF) V2(LF)

LK

ot-tate-ot-tate (< ofi-tate) 'keeping on chasing'

V1 (LF) + part of V1 (LF) + te

LK

utufusi-fusi-te 'staying prone a little bit'

The first two types are the reduplication of compound verb. The way to distinguish two types of functions, adverbial or linking, depends on the presence or absence of arguments and the semantic change of a given compound, as explained above. Consider the following examples. (11)

kore this

wo ACC

tot-te take-and

ni two

san three

ben times

osi-kafesi-osi-kafesi push(LF)-return(LF)-push(LF)-return(LF)

yomi-kik-ase... read-listen-CAUS

'(The ministeri) took this (i.e. letter) and made (the man) listen to (hisi) reading (it) repeatedly, twice or thrice...' (13C Heike, vol.1: 129) (12)

taka-domo eagle-PL

amata many

suwe-sase take-CAUS

udurafibari skylarks

ot-tate-ot-tate finemoso chase(LF)-stand(LF)-chase(LF)-stand(LF)

all.day

wo ACC ni PT

kari-kurasi hunt-spend

'(Sir Sukemori) made (his men) take many eagles, and (made the eagles) chase skylarks, and spent all day long hunting a n d . . . ' (13C Heike, vol.1: 77)

292

Reijirou Shibasaki

In (11), both the reduplicated verb osi-kafesi-osi-kafesi (lit. push-return) and the verb phrase yomi-kiku 'to listen to one's reading' are transitive in meaning; therefore, grammatically they are supposed to carry as the direct object the noun marked by the accusative case kore wo 'this (i.e. letter). However, if the former carried the noun as its direct object, the whole meaning would not fit this context: 'the minister pushed and returned this letter and made the man listen to his (i.e. minister) reading it'. Rather, the noun plays a direct object role of the latter verb in this context: 'the minister repeatedly made the man listen to his (i.e. minister) reading it'. In (11), the reduplicated compound ot-tate-ot-tate 'chase-stand-chase-stand' carries its direct object and chains into the following clauses; therefore, this is an example of linking verb. This form represents the reduplication of the compound verb ot-tatu (< ofu 'to chase' + tatu 'to stand'). The third type is different from all others above, because in it, the second part of the base (i.e. —fusi) is duplicated and then followed by the linking particle -te. In terms of type and token frequencies, these are not considered to be dominant among my examples (only two examples). Importantly, all these forms have linking forms; therefore, I will regard them as being a subgroup of V (LF) + V (LF) in this study.5

2.5. Repetition and reduplication Lastly, I will touch on the difference between repetition and reduplication which is applied to this study. Let us consider the following examples. (13)

hosi stars kara from

ga kagayai-te kagayai-te marude NOM twinkle(LF)-and twinkle(LF)-and as if ochi-te-ki-sou-de-at-ta fall-LK-come-INFER-COP-exist-PST

'Stars were twinkling so much as if (they) fell from the sky.' (Kaji 1993 Afurika ο Fiiirudowaakusurw. (14)

tabe-ta tabe-ta eat(LF)-PST eat(LF)-PST '(I) ate a lot.' (an example from my anonymous reviewer)

(15)

sake ο nomi-nomi alcohol ACC drink (LF)-drink(LF) '(I) thought while drinking alcohol.'

sora sk

60)

kangae-ta think-PST (Minami 1973:122)

On the grammaticalization of verbal reduplication in Japanese

293

In (13), the verb is suffixed by the linking particle -te and then duplicated, while in (14), the verb is suffixed by the past tense auxiliary and then duplicated. In other words, the verb phrases are duplicated after suffixation. In (15), however, neither of the duplicated verb forms is accompanied by any auxiliary or linking particle which often indicates modality; it only undergoes the morphological operation whereby the whole base nomi 'to drink' is copied and attached to the base to form nomi-nomi 'while drinking'. While these examples could be regarded as a different type of reduplication, I will regard (13) and (14) as repetition and (15) as reduplication respectively. It is true that the reduplicated compound verbs in Table 2 have also undergone suffixation. But I will regard them as a type of reduplication, because no linking particle is inserted and tense is unexpressed in much the same way as with other reduplication in Table 1. This criterion is not totally consistent with such approaches as Kageyama (1993: 264) and Okamoto (1994; cf. Okamoto 1990); however, I will refer to this criterion in this study to make sure the difference between repetition and reduplication.6

3.

Stages and Data

Approximate stages in the history of Japanese are given in (16). For each stage, I have carefully chosen those texts that are considered to include colloquial expressions of those times, as illustrated in (17). While it is mainly narratives which are to be analyzed, I will also analyze some diaries, essays and collections of verse songs. (16)

a. b. c. d. e.

Stage Stage Stage Stage Stage

1: 2: 3: 4: 5:

(17)

a.

Narratives Narratives:

Early Old Japanese (Early Ο J, 8C) Late Old Japanese (Late OJ, 9-12C) Middle Japanese (MJ, 13-15/16C) Pre-Modern Japanese (Pre-Mod J, 17-late 19C) Modern Japanese (Mod J, late 19-20C) Taketori (late 9C) Ise, Yamato, Heichuu, Otikubo (10C) Tutumichuunagon, Genji, vol.1 (11C) Ookagami (12C), Heike 1 & 2 (13C) Otogisausi, vol.2 (15/16C) Manjiban Isopo (17C) Tuyudono, Denbit, Motonomokuami (17C)

294

Reijirou Shibasaki b. c. d. e. f.

Diaries: Essay: Puppet Play: Fiction: Verse songs:

Izumishikibu (HC), Tofazugatari (14C) Turezuregusa (14C) Sayokoromo osidori no turugiha (18C) Gingatetudoo no yoru (20C) Man'yooshuu, vols. 1~14(&C)

In order to illustrate the diachronic proportional frequencies of verbal reduplication, I will examine types and tokens of verbal reduplication in Old through Modern Japanese. I will follow Hachiya (1998) and N. Yamaguchi (1984) for the categorization of reduplication unless otherwise specified. Verbal reduplication is considered to have undergone the paradigmatic change at stages 2 and 3 (Shibasaki 2001); therefore, I will survey texts particularly around those periods.

4.

Analysis

In this section, the diachronic transition of verbal reduplication is to be examined with focus on the two major types of verbal reduplication in Table 1; namely, V (SF) + V (SF) and V (LF) + V (LF). Various forms of each type are illustrated with their frequencies and examined to see what kind of verbal reduplication occurs with what number of tokens.

4.1. Types of V(SF) + V(SF) Table 3 shows the diachronic frequencies of eight types of V (SF) + V (SF) forms and reveals that by the end of MJ (13-16C), the verbal reduplication of V (SF) + V (SF) has almost been adverbialized in meaning. ADV=adverbial function, LK=clause-linking function. After Pre-Modern Japanese (17-19C), only some fossilized expressions have survived to Modern Japanese (20C). The proportional frequencies of linking and adverbial functions are summarized in Chart 1. This finding may be correlated with the gradual increase of ideophonic reduplication over history (Shibasaki 2001). As pointed out in Shibatani (1990: 155), many Japanese verbs indicate very general meanings in comparison to English. For example, the verb warau is a general term for laughing, which entails lack of specific manners as lexically indicated in the English verbs such as smile or chuckle.

On the grammaticalization

of verbal reduplication

in Japanese

295

Table 3. Diachronic frequencies of verbal reduplication 1 Function

Form V(SF) V(SF) V(SF) V(SF) V(SF) V(SF) V(SF) V(SF)

+ + + + + + + +

V(SF) V(SF) + V(SF) + V(SF) + V(SF) V(SF) + V(SF) + V(SF) +

mo to ni ιno to ni

LK LK LK LK ADV ADV ADV ADV

Total

8C

9-12C 13-16C 17-18C 20C

Total

3 0 1 3 3 0 0 7

7 2 1 0 54 1 11 0

1 0 0 0 36 3 10 1

0 0 0 0 9 0 4 0

0 0 0 0 3 0 0 0

11 2 2 3 105 4 25 8

17

76

51

13

3

159

Chart 1. Transition of linking to adverbial functions: V(SF) + V(SF)

oo

fN ON

Ό γί

oo r-

Ο tM

Furthermore, if we consider the Japanese event structure from the viewpoint of 'framing', Japanese is considered to realize the pattern that expresses motion along with paths (Talmy 2000: 49), while English is considered to realize the pattern that express motion occurring in various manners (Talmy 2000: 27). This observation tells us that even if a Japanese verb is reduplicated, it appears to be impossible to indicate specific meanings of manner; therefore, the gradual increase of ideophonic reduplication seems to make sense in that ideophones compensate for the lack of semantic expressiveness of the verb. Furthermore, the clause-linking function of verbal reduplication has been replaced by linking particles which began to occur frequently from Late OJ on (Konoshima 1966: 181-191; Hachiya 1998: 52-61). All in all, the overall frequency of verbal reduplication has decreased over time in all types of V (SF) + V (SF) forms.

296

Reijirou

Shibasaki

Table 4. Types of verbs Types of Verbs

Original Meaning

fqfu-fafu-(mo) ifofu-fofu-(to)) furu-furn kafesu-gafesu- (mo) karu-karu katu-gatu kawaru-gawaru kiru-kiru koro-goro-(ni) masu-masu-(ni)

to crawl

miru-miru misu-misu naku-naku niyofu-niyofu omofu-omofu siku-siku-(ni)

to see to see to cry to groan to think to happen L(2) sequentially A (5) to know to trace to attach to feel lonely to separate to sob

L(l) A (3) A (31) A (1) L(2)

to laugh to go

L(l)

A (1) L(2) A(l)

17

75

siru-siru tadoru-tadoru-(mo) tuku-duku-(to) wabu-wabu waku-waku wananaku -wananaku warafu-warafu yuku-yuku-(to)

Total

to to to to to to to to

fall return cut forebear change cut crowd increase

8C

9-12C

13-16C

17-18C

L(l) A (1)

L(l) A (1)

A(l)

A (14)

A (7)

A (2)

20C

5

L(l) L(l) A (1) A (3) L(l) L(l) L(l) A (4)

Total

1 23 1 1 3 1 1 6

A(l)

L(l) L(l) A (10) A (2) L(l) A (1)

A (29)

A (12)

1 3 65 1 2 7

A (5)

1 1 27 2 1 1

A (5)

1 4

51

13

3

159

The overall change in verbal reduplication by token frequency uncovers a directional pathway into adverb. However, since it is not yet clear whether this general tendency can be seen in all kinds of verbs in an equal way, I will illustrate the type frequency of verbal reduplication in Table 4. Notice that the main point here is to see whether a given verb has a tendency to become adverbial or not. To make it clearer, I will only focus on adverbial and linking functions without paying any special attention to subtypes of

On the grammaticalization of verbal reduplication in Japanese

297

forms such as V(SF) + V(SF) + to. L=linking function, A=adverbial function. Numbers in parentheses mean tokens of given verbs. Table 4 tells us that this general pathway of verb into adverb cannot equally be seen in every kind of verb; rather, it is skewed towards certain types. For example, emotion verbs have a strong tendency to be used adverbially, as in naku 'to cry', niyofu 'to groan', wabu 'to feel lonely', wananaku 'to sob', warafu 'to laugh'. All these verbs function adverbially once they are reduplicated from Early OJ on. Some of them can still be used in Modern Japanese with a certain old-fashioned taste (e.g. nakunaku), although my texts happened to include no such examples. Note that koro derives from korn (SF) through phonological change. Two perception verbs miru 'to see' and misu 'to see' also give insight into semantic change through reduplication. The reduplicated form mirumiru indicated 'while seeing (something)' in Late OJ, but it changed or expanded the meaning into 'in a moment' in later stages, though not included in my texts. While the other verb misu-misu was already used to render an adverbial meaning 'apparently' in Late Ο J, it came to indicate a concessive meaning 'despite knowing' in later stages. Both cases undergo unidirectional semantic change from verbal to adverbial. Some other clearer cases are to be investigated in detail in section 5, but suffice it to say here that the process of these semantic changes appears to be unidirectional. More importantly, Table 4 includes some crucial examples characteristic of grammaticalization. Although I will leave their detailed account until section 5, there are some cases where the newer and older functions coexist at a synchronic stage, i.e. 'layering' (Hopper 1991). In Early OJ, for example, when the verb masu 'to increase' was reduplicated, it came to render either 'to keep increasing' as a verb or 'more and more' as an adverbial. In Late MJ, however, the adverbial meaning came to predominate; both meanings are rare, but not impossible in Modern Japanese. Layering is not seen in all verbs in Table 4, but some clear cases may give support for this distributional change as a case of decategorization, a piece of evidence for grammaticalization. Lastly in this section, I will touch on cases of phonological change which is often found in the process of grammaticalization (Hopper and Traugott 1993: 1-2). Japanese is known to have sequential voicing or rendaku by its traditional name, which shows a voicing process in the initial consonant of the second form of a compound, unless there is already a voiced obstruent in this position (Shibatani 1990: 173-175). My observation is that if a given verb involves an obstruent in the initial consonant and its

298

Reijirou Shibasaki

reduplicated form becomes totally adverbial, sequential voicing is most likely to occur; in addition, only adverbialized reduplication undergoes sequential voicing. For example, the verb kawaru 'to change' undergoes sequential voicing when reduplicated as in kawaru-gawaru 'one after another', while the verb karu 'to cut' does not undergoes when reduplicated as in karu-karu 'cut (it) more'. The only exceptional case is the verb fafu 'to creep'. This verb changes not its initial consonant but its vowel when used adverbially through reduplication, as in fou-fou 'energetically' in (4). As Shibatani (1990: 175) points out, there are numerous cases that fail to undergo sequential voicing. When it comes to (verbal) reduplication, a given form tends to go through sequential voicing or some other phonological change, though not always, if it semantically becomes adverbial through reduplication. Therefore, the decategorization of verbal reduplication is considered to show phonological change typical in the process of grammaticalization.

4.2. Types of V ( L F ) + V(LF) In this section, I will delve into the other major formation of verbal reduplication: V (LF) + V (LF). All types of V (LF) + V (LF) are illustrated in Table 5. Notice that the main verb function by V(LF) + V(LF) + aux is counted as part of the linking function in Chart 2. ADV=adverbial function, LK= clause-linking function. Table 5. Diachronic frequencies of verbal reduplication 2 Form V(LF) + V(LF) + ni V(LF) + V(LF) + to V(LF) + V(LF) V(LF) + V(LF) V(LF) + V(LF) + te V(LF) + V(LF) + aux Total

Function ADV ADV ADV LK LK main V

8C 9-12C 13-16C 17-18C 20C

Total

3 2 0 0 6 0

3 0 1 0 15 1

13 6 1 0 6 4

0 3 3 2 1 0

1 0 0 0 0 0

20 11 5 2 28 5

11

20

30

9

1

71

In spite of the relatively smaller number of examples in comparison to Chart 1, the verbal reduplication of V (LF) + V (LF) has traced the same pathway from verbal to adverbial over time. An important difference is that

On the grammaticalization

of verbal reduplication

in Japanese

299

the linking function of V (LF) + V (LF) has survived into later stages than that of V (SF) + V (SF). As in Chart 1, the verbal reduplication of V (SF) + V (SF) accelerated the process of becoming adverbial in Late OJ (9-12C) and almost completed the decategorization by the end of MJ (13-16C). However, the verbal reduplication of V (LF) + V (LF) retained its linking function until Pre-Modern Japanese, as in Table 5 and Chart 2. Judging from Table 5, the construction with the linking particle -te appears to retain verbal properties of reduplication. Chart 2. Transition of linking and adverbial functions: V(LF) + V(LF)

oo



0\

oo

Ο

rA

Now let us look at the types of verbs in Table 6. Compared with V (SF) + V (SF) in Table 4, there seems to be nothing particularly skewed in the formation of V (LF) + V (LF). However, three crucial phenomena for grammaticalization, i.e. unidirectional semantic change, layering and phonological change, are involved in this formation too. For example, the verb aru 'to exist' renders the meaning 'to live long' when reduplicated in Early OJ, while it comes to indicate an adverbial meaning 'in the long run' in Pre-Modern Japanese. This semantic change appears to be unidirectional. Although my texts provide no evidence for the coexistence of linking and adverbial functions of this verb at a given synchronic stage, several dictionaries cite an adverbial function of this verb in Late OJ. Thus, this verb shows two pieces of evidence for grammaticalization; i.e. layering and unidirectional semantic change. The third criterion, phonological change, often occurs when verbal reduplication becomes adverbial. Most examples undergo phonological change, i.e. sequential voicing, as examined in section 4.1. Interesting is the case o f f o r e - f o r e - t o 'in an absent-minded way', which does not undergo sequential voicing in Early MJ despite its adverbial usage. In Pre-Modern

300

Reijirou

Shibasaki

Japanese, fore-fore-to goes through sequential voicing to form fore-bore-to while keeping its adverbial function. (The same is true of kure-kure-to > kure-gure-to 'in a gloomy mood' in Modern Japanese, although kure-gureto is not included in my texts.) Table 6. Types of verbs Types of Verbs

Original Meaning

ari-ari-te/-to to exist fore-fore-to to be out (or fore-bore-to) of mind to draw furi-ide-ide-zo huruhi-huruhi-zo to swing ifi-ifl-te to say to combine kane-gane kari-(nomi)-kari- te to cut kire-gire-ni to cut to disappear kiye-kiye-to kofl-kofl-te to love kure-kure-to to get dark miti-miti-te to abound nami-nami-ni to line up nare-nare-te to get used to nozoki-nozoki to peek to get wet nure-nure-to omofl-omoft-(ni) to think sime-jime-to to soak simi-jimi-to to soak sinobi-sinobi-ni to adore tafe-dafe-(ni) to extinct tadori-tadori-to to trace tanome-tanome-te to ask tiri-diri-(ni) to disperse tori-dori-(ni) to take tugi-tugi-(ni) to connect ture-dure-ni to follow utufusi-fusi-te to stay prone utufusi-utufusi to stay prone -ni-keri wake-wake-te to separate yuki-yuki-te to go Total

8C

L(l)

9-12C

13-16C

17-18C

A(l)

A (1) A (2)

L(l)

20C

3 1

LCD

A (1) L(6) A (1) L(l) A (1) A(l) L(3) A(l) L(9) A(l) L(l) L(2) L(2)

A(l) A(l)

A(l) A A A A

(2) (1) (3) (1)

A (1)

A (1)

L(l)

A (2)

A(l) A (2) A(l) L(2) L(l)

A (6) A (3)

A (1)

L(l) L(l)

L(2)

11

20

30

Total

1 1 6 1 1 1 1 3 1 9 1 1 2 1 3 1 3 1 3 2 1 7 4 4 1 2 1 1 3

9

1

71

On the grammaticalization of verbal reduplication in Japanese

301

The case of kiye-kiye-to 'in a lifeless mind' is also crucial for grammaticalization. Although only kiye-kiye-to is involved in my texts, some dictionaries cite kiye-giye-to at the same synchronic stage, and both are adverbial functions. This gives evidence for the layering of examples of this verbal reduplication in MJ, and only ki(y)e-gi(y)e-to is used in Modern Japanese. 7 These cases imply that the process of grammaticalization is not abrupt but gradual. Grammaticalization theory tells us that grammaticalization is often accompanied by layering, phonological change and unidirectional semantic change. But, the above examples show that those properties do not necessarily occur at the same time. Furthermore, such a frequencybased observation gives insight into the possibilities of the same development of other verbal reduplication. In other words, even though there is no layering in all examples in Tables 4 and 6, the same process of grammaticalization would certainly have progressed in other texts.

4.3. Overall transition of verbal reduplication In this section, I will illustrate the overall transition of linking and adverbial reduplication, combining Tables 3 and 5. Table 7 shows the tokens of linking and adverbial functions, and Chart 3 summarizes the percentages of each function. Table 7:

Overall frequencies of verbal reduplication

Function

8C

9-12C

Clause-linking Adverbial

13 15

25 70

Total

28

95

13-16C

17-18C

20C

Total

11 70

3 19

0 4

52 178

81

22

4

230

Proportional frequencies of both functions have been changing steadily since Early OJ (8C). The linking function has decreased to zero in Modern Japanese (20C), at least in my texts. While the adverbial function has dramatically increased to the present in terms of proportional frequencies, the token frequencies have decreased almost to zero. As I explained in section 4.1, the emergence of several linking particles has gradually taken over the

302

Reijirou

Shibasaki

linking functions of verbal reduplication. I will not examine this phenomenon any further, but suffice it to say that the grammatical ization process has come almost to an end in Modern Japanese. The important aspect of this grammaticalization is that before Modern Japanese, the linking and adverbial functions of verbal reduplication have co-existed at each synchronic stage despite their distributional differences. Previous research has paid no special attention to the functional shift by which verbal reduplication has lost its verbal properties. However, both the co-existence and the steady decline of the two functions in reduplicative forms over history are suggestive of the nature of language change, i.e. grammaticalization. Chart 3. Overall proportional frequencies of linking and adverbial functions 100%

• Link HAdv

This phenomenon is pervasive across languages. For example, I found no evidence of adverbialized reduplication developing into lexical or linking verbs in a cross-linguistic study of this phenomenon in East and Southeast Asian languages (Shibasaki 2001). Michael (1970: 73) assumes that if lexical items lose their own functions, they would go along a directional pathway into adverb. In African languages, though not necessarily with verbal reduplication, verbs are often used where English speakers might use adverbs (Lord 1993: 215-233). Therefore, the development of verbal reduplication into adverbial phrases appears to be common.

On the grammaticalization of verbal reduplication in Japanese 5.

303

Two case studies

I have thus far taken a macro-scopic perspective to give a broad outline of Japanese verbal reduplication. In this section, I will take a micro-scopic perspective to summarize how grammaticalization has progressed in verbal reduplication by examining two clearer cases in detail.

5.1. Case of yuku-yuku My texts include four examples of yuku-yuku or yuku-yuku-to, and the layering of linking and adverbial functions is found in Late OJ. The original meaning 'to go' in yuku comes to render 'while going' when reduplicated as early as Early OJ, as in (18), and the same function is also found in Late OJ, as in (19). (18)

Nifu-no-kawa the.Nifu.river

se wa rapids TOP

watara-zu-te yuku-yuku-to... cross-NEG-and go(SF)-go(SF)-PT

'While (I) am going along without crossing the rapids of the Nifu river...' (8C Man 'yoo: 130) (19)

kimi you

ga NOM

sumu live

yado house

no GEN

kozuwe treetops

wo ACC

yuku-yuku-to... go(SF)-go(SF)-PT 'While (I) am going along the treetops in the house you live...' (12C Ookagami: 92) Importantly, the reduplicated verb yuku-yuku 'while going' co-occurs with locative expressions, which may function to retain the verbal property of the process verb yuku. On the other hand, yuku-yuku comes to occur without any locative expression at the same synchronic stage, as in (20). Notice that in (20), the particle -to detaches from yuku-yuku. (20) yuku-yuku go(SF)-go(SF)

nomi-kufu drink-eat

'(We) ate and drank while going.'

(10C Tosa\ 14)

304

Reijirou Shibasaki

The adverbial function of yuku-yuku also emerges in Late OJ, as in (21). In this example, the reduplicated verb yuku-yuku indicates 'quickly, smoothly' without carrying any locative expression, whereby the original meaning of motion shifts into a new meaning of manner (cf. Hachiya 1998: 332). Notice again that the adverbialized yuku-yuku does not take the particle -to any longer; therefore, the form and function of yuku-yuku 'while going' or 'quickly' is considered to be in transition under grammaticalization. (21)

nani what

goto thing

ni PT

to COMP

miya Princess

kafa QP ni PT

todokofori-tamafu take.time-HON mo also

yuku-yuku go(SF)-go(SF)

urefe-kikofe-tamafu complain-HON-HON

'"Why (do you) take time, (do it) quickly.", (the minister) complained (even) to the Princess.' (11C Genji, vol.1: 411 —412) From Pre-Modern Japanese to the present, yuku-yuku further shifts its adverbial meaning to indicate 'in the course of time, near future', as cited in several dictionaries. Example (22) from Modern Japanese shows the new meaning, although it is a constructed example based on one consultant's judgment. (22) yuku-yuku wa kekkon go(SF)-go(SF) TOP marriage '(We) will marry in the near future.'

suru do

tsumori-desu will-POL

In Modern Japanese, however, the other two meanings 'while going' and 'smoothly' seem to have fallen out of use. Now I will summarize the diachronic development of yuku-yuku in Diagram 1. The examples of yukuyuku indicating 'smoothly' and 'while going' in MJ are added from Hachiya (1998: 101-106).

On the grammaticalization of verbal reduplication in Japanese

305

Diagram 1. Grammaticalization ofyuku-yuku Form-and-Meaning yuku-yuku-to 'while going' yuku-yuku 'while going' yuku-yuku 'smoothly' yuku-yuku 'in the course of time'

Early OJ

Late OJ

MJ

Pre-Mod J

Mod J

^

^ ^

Two crucial properties of grammaticalization are involved in this diagram. One is the layering of linking and adverbial functions in Late OJ. Both yuku-yuku-to and yuku-yuku function to mean 'while going', which is a morpho-syntactic layering, while yuku-yuku renders either the meaning of 'while going' or 'smoothly', which is a semantic layering. The form yukuyuku shows a decategorization from verbal to adverbial through morphosyntactic change in Late OJ, and it comes further to indicate a new adverbial meaning 'in the course of time' while keeping its form from MJ on. This process appears to show a type of grammaticalization. The other crucial property concerns unidirectional semantic change. Firstly, the original process meaning 'to go' comes to mean 'while going' through reduplication. The process of 'going' tends to render an inference of futurity from a cross-linguistic perspective (Heine and Kuteva 2002: 161-163); therefore, the adverbial meaning 'in the course of time' inherits its motion or process meaning from 'while going', changing the concrete motion into an abstract process, as in Diagram 1 (cf. Hopper and Traugott 1993: 2 - 4 for the grammaticalization of be going to). Secondly, the continuous progress of 'going' may have been related to its smooth progress: 'while going (on)' > 'smoothly'. While this semantic extension has not been reported in works on grammaticalization (e.g. Heine and Kuteva 2002), it deserves investigation as a type of grammaticalization, especially, with respect to verbal reduplication. Finally, the semantic relation between 'smoothly' and 'in the course of time' could be explained by the original meaning of yuku 'going' which is most likely to be goal-oriented (e.g. andative or purposive) across languages. Note that sequential voicing is not involved in this grammaticalization, because the first consonant of the second form is originally voiced.

306

Reijirou Shibasaki

5.2. Case o f f a f u - f a f u The verb fafu is a motion verb to indicate 'to crawl'. In Late OJ, the reduplicated form fafu-fafu came to be used to mean 'while crawling', as in (23). Note that the particle mo often attaches onto V (SF) + V (SF), rendering a concessive meaning (Hachiya 1998: 57; cf. Konoshima 1966: 297299). (23)

fafu-fafu mo kimi ga atari ni sitagafa-mu crawl(SF)-crawl(SF) PT you GEN side LOC serve-will 'Even if (I have to) crawl (towards you), (I) will wait on you.' (11C Tutumichuunagon: 448)

In the same synchronic period, the reduplicated form creates a new meaning 'energetically' through phonological change and reduction (fafu > fou). (2) is repeated as (24) below. (24)

kasa umbrella

wo fou-fou-to ACC crawl(SF)-crawl(SF)-AP

no NOM/GEN

ito very

ookaru be.many

ufe on

ni in

ute-ba kuso hit-because dung

kagamari stoop

wi-nu sit-PERF

'Because (the messenger) brandished the umbrella energetically, (the two men) sat down on the dirty place.' (10C Otikubo: 124) In (24), the reduplicated form fou-fou functions adverbially to modify the verb ute 'to hit' (< utu). Suppose that fou-fou still remains a lexical verb, the whole clause makes no sense in this context: 'the messenger brandished the umbrella while crawling'. These two examples show a clear concomitant development of linking and adverbial functions o f f a f u and fou, i.e. layering. In MJ, these two functions still continue to be layered. The linking function is illustrated in (25), but another new adverbial meaning 'barely' emerges in this stage, as in (26). (25)

kefunisite scarcely

tasukari-taru survive-NM

fafu-fafu crawl(SF)-crawl(SF)

ie house

sama manner

nite PT

ni LOC

iri-ni-keri enter-PERF-PST

'(The priest) looks like having scarcely survived (and) entered (his) house while/by crawling.' (14C Turezure: 152)

On the grammaticalization of verbal reduplication in Japanese 26)

Sukemori-asomi fafu-fafu proper.name-minister creep(SF)-creep(SF)

307

rokufara fe place.name LOC

ofasi-te... come-and 'The minister Sukemori barely arrived in Rokufara and...' (13C Heike vol.L·. 78) According to Nippojisho, the dictionary published by the Portuguese missionaries in 1603, fafu-fafu is regarded as an adverb that is equal to yau-yau 'gradually, barely'. My texts include one example of hou-hou ([f] > [h]; see note 1) in Pre-Modern Japanese, as in (27). (27)

Sonogo Shishiou hou-hou-to oki-agari... after.that Lion.King creep(SF)-creep(SF)-AP rise-rise 'After that, the Lion King barely sat up and...' (17C Manjiban Isopo-monogatari: 111)

Several dictionaries cite no such examples after Pre-Modern Japanese; therefore, more philological research is needed to uncover the whole possible process of this grammaticalization. I will tentatively summarize the diachronic process o f f a f u - f a f u in Diagram 2. Diagram 2. Grammaticalization o f f a f u - f a f u Form-and-Meaning

Early OJ

Late OJ

MJ

Pre-Mod J

Mod J

fafu-fafu 'while creeping' fou-fou 'energetically' fafu-fafu 'barely' hou-hou-to 'barely'

The meaning o f f a f u 'to crawl' may be paraphrased into 'to progress vigorously but with difficulty (like bugs or a person with a load) along a surface in a crouching position'. The interesting thing is that the manner meaning connoted in fafu, i.e. 'vigorously', first shifted into 'energetically' in the

308

Reijirou Shibasaki

reduplicative form fafu-fafu in much the same way as yuku-yuku in the previous section. The emergence of the other adverbial meaning 'barely' can be accounted for in the following way. As in the previous section, verbs of motion or process tend to be goal-oriented (e.g. andative or purposive) across languages and probably the same applies to fafu. On the other hand, the manner meaning 'vigorously' may have been integrated into the goaloriented meaning. As a consequence, an inference of futurity based on the two meanings may have emerged as 'barely (achieve one's goal)', which may still reflect the other manner meaning 'with difficulty' of fafu. Of course, more research is required on this semantic shift, but both layering and unidirectional semantic change are involved in this case. Moreover, a phonological change and reduction {fafu > fou > hou), though not sequential voicing, is also accompanied by the change from verbal to adverbial functions in this case. In this whole section, I have examined the two cases that involve properties characteristic of grammaticalization. As illustrated in Tables 4 and 6, my texts do not necessarily include crucial examples that are supportive of grammaticalization. But, these two clearer cases are suggestive of the possibility of the same development of other verbal reduplication in Japanese and other languages.

6.

Concluding remarks

In this study, I have given a broad outline of the diachronic status and transition of Japanese verbal reduplication. The layering of verbal and adverbial functions has been witnessed at each synchronic stage, while changing their distributional frequencies over history. Japanese verbal reduplication has followed a unidirectional path in the grammaticalization of verb into adverbial. Once a given reduplicative verb becomes adverbialized in meaning and function, it undergoes phonological change, i.e. sequential voicing, more often than not. It is true that this small study is just a preliminary step for a fuller analysis of Japanese verbal reduplication. However, findings from this diachronic study will reinforce or reinterpret those from crosslinguistic studies. It is hoped that a combination of diachronic and crosslinguistic studies will further uncover what is yet to be uncovered in the languages of the world.

On the grammaticalization

of verbal reduplication

in Japanese

309

Notes *

1.

2.

3.

4.

5.

Especial thanks to Susanna Cumming, John W. Du Bois, Carol Genetti and Sandra A. Thompson and an anonymous reviewer for comments on this paper and insights that have contributed directly to it; also to Bernard Comrie, Yuriko Sunakawa and Hideki Tsukamoto for the last minute input into it. The remaining faults are all my own. The sounds described by [f] from the 8 th through 16th centuries indicate the voiceless bilabial fricative [φ ]. I used [f] just for convenience in this study. The voiceless bilabial fricative becomes the voiceless glottal fricative [h] after the 17th century (e.g. Komatsu 1981: 249). The glossing conventions are: ACC=accusative; ADV=adverbial function; AP=adverbial particle; aux=auxiliary; CAUS=causative; COMP=complementizer; CONT=continuative; COP=copula; DAT=dative; DEM=demonstrative; DIST=distributive; EMPH=emphatic; FREQ=frequentative; GEN=genitive; HAB=habitual; HON=honorific; IMPERA=imperative; IMPERF=imperfective; INFER=inferential; IR=irrealis; LF=linking form; LK=linking form; LOC=locative; NEG=negative; NM=noun-modifying form; NOM=nominative; PERF=perfective; POL=polite; POTEN=potential; PROG=progressive; PST=past; PT=particle; SF=sentence-final form; SPON= spontaneous; TOP=topic; V=verb. Unexpressed elements (e.g. subjects) are put in parentheses. Typologically, this is called 'anterior clause chaining' (Haspelmath 1995: 22). In Japanese, it is often difficult to make a clear distinction between tense and aspect. For example, the past tense auxiliary -keri is considered to have derived from ki-ari (come-exist) 'has just come' (Kasuga n.d. cited in Y. Yamaguchi 1985: 511-518). However, the sentence-final form of the verb basically indicates the present tense and indicative mood (Y. Yamaguchi 1985: 492; Teramura 1984: 75-113). The other formation is reported in Hachiya (1998: 85-86) and consists of linking and realis forms such as yuki-yuke-ba 'when (I) went along'. Note that the particle -ba can be connected either to irrealis or realis forms of the verb. There are two other formations which are sometimes considered to be reduplication or syntactic idioms: verb-«/-verb, verb-to-verb (Yamada 1954: 453; Konoshima 1966: 82; Kageyama 1993: 89-92; Okamoto 1990; cf. 1994; Miyaoka 2002: 92 nt. 34). In these formations, the linking particles -ni and -to combine two verbs both of which are linking forms. I will not discuss these formations in this study, but they have some semantic properties in common with the above examples. I owe this observation to the anonymous reviewer.

310

Reijirou

Shibasaki

6. Ishikawa (1991: 565, 576) argues, for example, that reduplication shows a correspondence between form and semantic meaning, while repetition shows a correspondence between form and interactional function in discourse. 7. Verbs consisting of more than two morae seem to undergo no sequential voicing when reduplicated, as in huruhi-huruhi-zo. Further research is needed.

References (1) Primary sources Fujioka, Tadaharu, Koichi Nakano, Kiyoshi Inukai, and Fumio Ishii (eds.) 1994 Izumishikibu-nikki, Murasakishikibu-nikki, Sarashina-nikki, and Sanukinosuke no nikki. Tokyo: Shogakkan Ichiko, Teiji (ed.) 1973 Heike-monogatari 1 (vol.1-6). Tokyo: Shogakkan. 1986 Otogizoushi vol.2. Tokyo: Iwanami bunko. Jin'pou, Kazuya, Tadakazu Aoyama, Tokuzou Kishi, Masachika Taniwaki, and Tsuyoshi Hasegawa (eds.) 1971 Kanazoushi and Ukiyozoushi. Tokyo: Shogakkan. Kaji, Shigeki 1993 Afurika ο Fiiirudowaakusuru. Tokyo: Taishuukan Shoten. Katagiri, Youichi, Teisuke Fukui, Shoji Takahashi, and Yoshiko Shimizu (eds.) 1972 Taketori-monogatari, Ise-monogatari, Yamato-monogatari, and Heichuu-monogatari. Tokyo: Shogakkan. Kojima, Noriyuki, Masatoshi Kinoshita, and Akihiro Satake (eds.) 1971-73 Man'yoosyuu 1-3 (vol. 1-14). Tokyo: Shogakkan. Mitani, Eiichi and Keiji Inaga (eds.) 1972 Otikubo-monogatari and Tutumichuimagon-monogatari. Tokyo: Shogakkan. Miyazawa, Kenji 1961 Gin 'gatetsudoo no yoru. Tokyo: Shincho bunko. Muto, Sadao (ed.) 2000 Manjiban Isopo-monogatari. Tokyo: Iwanami-shoten. Tachibana, Kenji (ed.) 1974 Ookagami. Tokyo: Shogakkan. Tsunoda, Ichiro and Uchiyama, Mikiko (eds.) 1991 Takeda Izumo, Namiki Sosuke Joorurisyuu. Tokyo: Iwanami-shoten. Yamagishi, Tokuhei (ed.) 1958 Gen'ji-monogatari I. Tokyo: Iwanami-shoten. Yasuraoka, Kosaku (ed.) 1971 Tsurezuregusa. Tokyo: Obunsha.

On the grammaticalization

of verbal reduplication

in Japanese

311

(2) Secondary sources Bybee, Joan, Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Bybee, Joan and Sandra A. Thompson 2000 Three frequency effects in syntax. Proceedings of the twenty-third annual meeting of the Berkeley Linguistics Society: 65-85. Hachiya, Masato 1998 Kokugochoofukugo no Gokooseironteki Kenkyuu [A word formational study in Japanese reduplication], Tokyo: Hanawa Shoboo. Hagege, Claude 1993 The Language Builder. Amsterdam: John Benjamins. Haspelmath, Martin 1994 Converb as a cross-linguistically valid category. In Converbs in Cross-Linguistic Perspective, Martin Haspelmath and Ekkehard König (eds.), 1-55. Berlin/New York: Mouton de Gruyter. 2002 Understanding Morphology. London: Arnold. Heine, Bernd and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hinton, Leanne, Johanna Nichols, and John Ohara (eds.) 1994 Sound Symbolism. Cambridge: Cambridge University Press. Hopper, Paul J. 1991 On some principles of grammaticalization. In Approaches to Grammaticalization, vol.1, Elizabeth C. Traugott and Bernd Heine (eds.), 17-35. Amsterdam: John Benjamins. Hopper, Paul J. and Elizabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Ishikawa, Minako 1991 Iconicity in discourse: The case of repetition. Text 11 (4): 553-580. Kageyama, Taro 1993 Bumpoo to Gokeisei [Grammar and Word Formation]. Tokyo: Hituzi Shoboo. Kasuga, Seiji n.d. Saidaiji Kinkoomyoo Saishoooukyoo Kuten no Kokugogakuteki Kenkyuu [A linguistic study in the punctuation system in Kinkoomyoo Saishoooukyoo, Saidaiji version]. Komatsu, Hideo 1980 Nihon'go no Sekai 7: On'in [The Universe of Japanese 7: Phonology], Tokyo: Chuuookooronsya. Konoshima, Masatoshi 1966 Joshi Jodooshi no Kenkyuu [Studies in particles and auxiliaries]. Tokyo: Oofuusha.

312

Reijirou

Shibasaki

Longacre, Robert Ε. 1985 Sentences as combinations of clauses. In Language Typology and Syntactic Description, vol.11, Complex Constructions, Timothy Shopen (ed.), 235-286. Cambridge: Cambridge University Press. Lord, Carol 1992 Historical Change in Serial Verb Constructions. Amsterdam: John Benjamins. Michael, Ian 1970 English Grammatical Categories. Cambridge: Cambridge University Press. Minami, Fujio 1973 Gendai Nihin'go no Koozoo [The structure of Modern Japanese], Tokyo: Taishuukan Shoten. Miyaoka, Osahito 2002 Go towa Nani ka [What Word is]. Tokyo: Sanseidoo. Moravcsik, Edith 1978 Reduplicative constructions. In Universals of Human Language, vol. Ill, Word Structure, Joseph Greenberg (ed.), 297-334. Stanford, CA: Stanford UP. Okamoto, Shigeko 1990 Reduplicated verbs in Japanese as grammatical constructions. Proceedings of the sixteenth annual meeting of the Berkeley Linguistics Society. 248-256. 1994 Augmentative verbal repetitive constructions in Japanese. Cognitive Linguistics 5 (4): 381-404 Shibasaki, Reijirou 2001 Construction reanalysis and paradigmatic change in Japanese reduplications. Paper presented at the 7th International Cognitive Linguistics Conference, at the University of California, Santa Barbara, July, 2001. Shibatani, Masayoshi 1990 The Languages of Japan. Cambridge: Cambridge University Press. Syromiatnikov, N. A. 1981 The Ancient Japanese Language. Moscow: Nauka: Publishing House. Talmy, Leonard 2000 Toward a Cognitive Semantics, vol.2: Typology and Process in Concept Structuring. Cambridge, Mass.: MIT Press. Teramura, Hideo 1984 Nihon 'go no Shintakusu to Imi 2 [Syntax and semantics of Japanese, vol. 2]. Tokyo: Kuroshio Publishers.

On the grammaticalization

of verbal reduplication

in Japanese

313

Yamada, Yoshio 1936 Nippon Bumpoo Koogi [Lectures on the Japanese grammar]. Tokyo: Hoobunkan. Yamaguchi, Nakami 1984 Heianbun'gaku no Buntai no Kenkyuu [Studies in stylistics in the Heian era], Tokyo: Meiji Shoin. Yamaguchi, Yoshinori 1985 Kodai Nihon'go Bumpoo no Seiritsu no Kenkyuu [Studies in the birth of Old Japanese grammar], Tokyo: Yuseido.

Reduplicative allomorphy and language prehistory in Uto-Aztecan Jason D. Haugen

1. Introduction 1 Although Uto-Aztecan is one of the largest and most well-established language families of North America, the internal organization of sub-groups within Uto-Aztecan is still a matter of debate. At the center of discussion is whether or not the geographical split of Uto-Aztecan into "Northern" and "Southern" branches can also be considered to be a genetic split. Also at issue is the legitimacy of various proposed sub-groups within these two larger possible branches. The Northern branch is generally considered to be a legitimate unit, based on shared morphological features (Heath 1977, 1978) and a sound law involving the lenition of intervocalic **-c- to *-y(Manaster Ramer 1992). The unity of the Southern branch is more controversial but it does have its proponents, largely due to lexicostatistical evidence (i.e. percentage of shared cognates) (Miller 1984; Cortina-Borja and Valinas 1989). In addition, for some sound correspondences between the two branches, such as NUA /q/ ~ SUA /n/ and NUA /η/ ~ SUA /l, r/, it is difficult to determine which branch is the innovator, with some scholars (e.g. Hill 2001) arguing for innovation in NUA, and others (e.g. Silver and Miller 1997) for innovation in SUA. 2 Nevertheless, for the reconstruction of grammatical elements back to the level of Proto-Uto-Aztecan (PUA), it is necessary to provide corresponding elements from both NUA and SUA as a minimal requirement for a likely PUA reconstruction. Following the call to find evidence beyond shared sound changes in the establishment of genetic relationships heralded by such scholars as Thomason and Kaufman (1989), Campbell (1997), and Dixon (1998), among many others, this paper began as an attempt to discover patterns of synchronic reduplicative allomorphy in an effort to identify shared allomorphy patterns, and thus possible family-internal relationships, among the languages within Uto-Aztecan. Barragan and Haugen (2002) identify four reduplication patterns in the Uto-Aztecan languages of Miller (1984)'s pro-

316

Jason D. Haugen

posed "Sonoran" branch of Southern-Uto-Aztecan, and in this paper I present evidence that the same patterns appear across Uto-Aztecan, in both the Northern and Southern branches, and can thus be reconstructed for the protolanguage. These reduplication patterns are a light syllable reduplicant; a disyllabic (CV(C)V-) reduplicant (or possibly a full root reduplicant); a bare mora affix; and a fourth reduplication type which surfaces as a heavy syllable variation on the light syllable reduplicant, and which will be called here "marked heavy syllable reduplication". All of these patterns of reduplication are retained in at least one UtoAztecan language: Yaqui.3 While it is true that the typology of reduplication forms is limited cross-linguistically, the occurrence of all of these patterns in this language and others of the family is more likely to be the result of shared inheritance rather than convergent development in multiple languages. These varying reduplication patterns would provide evidence for sub-groupings within the family if it could be demonstrated that certain patterns were limited only to certain languages, but the actual distribution of the patterns in both the Northern and Southern branches does not allow for any particular sub-groupings. That is, the lack of evidence for shared innovation in the domain of reduplication in sub-groupings of Uto-Aztecan suggests that each of these patterns must have existed in PUA. This paper is structured as follows. Section 2 will illustrate each of the reduplication patterns as they surface in one Uto-Aztecan language, Yaqui, and Section 3 will then provide examples of cognate reduplication patterns in both the Northern and Southern branches of Uto-Aztecan. Section 4 will discuss previous efforts at the historical reconstruction of Uto-Aztecan reduplication patterns and the reconstruction proposed here for PUA, and section 5 will discuss the issue of diachrony in reduplication more generally. Section 6 will conclude.

2. Reduplicative allomorphy in Yaqui (SUA) Spaelti (1997) distinguishes between two different kinds of multi-pattern reduplication. The first type consists of "duplemes", where different reduplication patterns (i.e. different reduplicative morphemes) are used for different semantic functions. An example would be the three patterns used in Sawai (Whistler 1992, cited by Spaelti 1997), where one reduplicant copies only the first consonant of the base for the durative (e.g. gslay 'to scream' ge-gelay 'wailing'), a second copies both the first and second consonant for nominalization (e.g. lesen 'to sweep' -> les-lesm 'broom'), and a third

Reduplicative allomorphy and language prehistory in Uto-Aztecan

317

copies only the second consonant for the reciprocal (e.g. gali 'to help' fal-gali 'to help one another'). The second kind of multi-pattern reduplication is the "alloduple", where different patterns of reduplication serve the same semantic function, but their shapes are phonologically conditioned. An example of this type is the Ponapean durative, where the reduplicant surfaces in a variety of shapes, depending on the phonological properties of the base: (1)

Ponapean durative (Rehg 1981) a. tep tepi-tep *ten-tep b. dod -> don-dod *dodi-dod c. pa -> paa-pa

'beginning' 'frequenting' 'weaving'

Spaelti acknowledges the existence of mixed systems, of which the Southern Uto-Aztecan language Yaqui is apparently a case: different (notentirely-prosodically-determined) patterns can express the same function (e.g. habitual action), and the same pattern can indicate different meanings. I use the term reduplicative allomorphy in a theory-neutral way to encompass both kinds of multiple-pattern reduplication within a mixed system. I assume that the mixed system in Yaqui has come about via convergence, where what used to be different duplemes have merged through time, giving the synchronic effect of suppletive allomorphy where multiple reduplicant shapes serve the same semantic function (usually habitual action). The facts of Yaqui reduplication are discussed in more detail in Haugen (2003), which discusses the phonology of the allomorphy in Yaqui reduplication, as well as in Harley and Amarillas (2003), which gives a detailed account of the various semantic functions of reduplication in Yaqui. The Yaqui data in this paper come from Molina, Valenzuela, and Shaul (1999), or forms collected in the Yaqui Reduplication Survey at the University of Arizona. Reduplication can serve several purposes within Yaqui grammar, and reduplication can co-occur with verbal tense/aspect/mood suffixes, nouns, numerals, adjectives, etc. The focus here (and throughout this paper) will be on prefixal reduplication, which usually expresses habitual action in Yaqui. However, as Harley and Amarillas show, each of the reduplication patterns in Yaqui can also be used to mean other things as well: iterative, continuative, imperative, etc. The semantics of the Yaqui reduplicant are often tied to the meaning of the verb root. Yaqui is a pitch accent language, and pitch accent does not determine the shape of the reduplicant. Thus, there are minimal pairs such as bwi.chi.a 'worm' and bwi.chi.a 'smoke', both of which reduplicate as bwLbwi.chi.a, which can mean either 'wormy' or 'smoky'.

318

Jason D. Haugen

On the received view of Yaqui reduplication (e.g. Escalante 1985; Martinez 1995; Demers, Escalante, and Jelinek 1999, etc.), there are two kinds of reduplication in Yaqui: "primary reduplication" with a light syllable reduplicant that copies the first syllable of the base (e.g. bwii.ka 'sing' bwi.bwi.ka 'usually sings'), and "secondary reduplication" with a heavy syllable reduplicant that triggers gemination of the onset of the base into coda position of the reduplicant (e.g. bwii.ka 'sing' bwib.bwi.ka 'sing from time to time'). We will see that this does not completely capture the facts of Yaqui reduplication, since there are two more kinds of habitual reduplication: disyllabic (or full root) reduplication and mora affixation (usually appearing as morphological gemination). In sum, Yaqui has three (largely unpredictable) "alloduples" of a single dupleme, 'habitual action': a light syllable reduplicant, a disyllabic reduplicant, and a bare-mora affix. In addition, there is another dupleme which surfaces as a phonological twist on the light syllable reduplicant: the "marked heavy syllable reduplicant." All four of these patterns have cognates across the Uto-Aztecan language family.

2.1. Yaqui light syllable reduplication The usual pattern of reduplication in Yaqui is light syllable reduplication (previously "primary reduplication"). This pattern exhibits a kind of reduplication that has often been claimed not to exist: the pattern of "syllable copy" reduplication (cf. Moravcsik 1978, Marantz 1982, McCarthy and Prince 1986, etc.). This reduplicant usually appears as a copy of the entire first syllable of the base: (2)

Yaqui monosyllabic habitual reduplication (Molina et al. 1999) a. b. c. d. e.

vu.sa chi.ke he.wi.te ko.'a.rek cho.'i.la

'awaken' 'comb one's hair' 'agree' 'wear a skirt' 'lasso'

vu.vu.sa chi.chi.ke he.he.wi.te ko.ko.'a.rek cho.cho'ila

a. b. c. d. e.

vam.se chep.ta chuk.ta hit.ta bwal.ko.te

'hurry' 'jump over' 'cut with a knife' 'make a fire' 'soften, smooth'

vam.vam.se chep.chep.ta chuk.chuk.ta hit.hit.ta bwal.bwal.ko.te

Reduplicative allomorpky and language prehistory in Uto-Aztecan

319

There are a few cases where a coda consonant does not copy (e.g. bwakta bwa.bwakta 'take out of a container'), but no cases where we see copy into a second syllable: i.e. there are no forms like *vus.vusa. Because the reduplicant surfaces with a short vowel, Demers et al. (1999) and Haugen (2003) argue that the reduplicant is a light syllable, and coda consonants are not moraic, except when they are geminate consonants serving some morphological function. Also, Haugen (2003) argues that the base for reduplication in these cases is only the first syllable of the word.

2.2. Yaqui marked heavy syllable reduplication Previously referred to as "secondary reduplication", the marked heavy syllable reduplicant surfaces as a heavy syllable reduplicant which triggers gemination from the onset of the base into the coda position of the reduplicant: (4)

Yaqui marked heavy syllable reduplication (Molina et al. 1999) a. b. c. d. e.

bwii.ka tee.ka va.hu.me ye.na 'om.te

'sing' 'lay it across' 'swim' 'smoke (tobacco)' 'get angry'

bwib.bwi.ka tet.te.ka vav.vahu.me yey.ye.na o'.'om.te

The marked heavy syllable reduplicant is always a different dupleme from habitual reduplication, and because it is always a different morpheme it is an example of what Urbanczyk (this volume) has referred to as the "enhancing of contrast" between morphemes. However, this process is apparently losing its productivity in the Arizona dialect, with speakers varying as to whether, and if so to what extent, this process is active. In many cases, the semantics of this reduplication pattern are idiosyncratic to the verb root. This loss of productivity is probably a concomitant of language shift in the Arizona Yaqui community.

320

Jason D. Haugen

2.3. Yaqui disyllabic reduplication Disyllabic habitual reduplication copies the first CV(C)V- sequence of the word: (5)

Yaqui disyllabic habitual reduplication (Molina et al. 1999) a. b. c. d. e.

ha.rah.te ha.soh.te ka.muk.ta ki.nak.te ku.pik.te

'crack lips' 'breathe hard' 'take a drink' 'squint, grimace' 'blink eyes'

ha.ra.ha.rah.te ha.so.ha.soh.te ka.mu.ka.muk.te ki.na. ki.nak.te ku.pi.ku.pik.te

a. b. c. d.

he.chi.te he.o.ta mo.i.na ri.u.ta

'scratch' 'wipe self' 'shoot' 'fracture'

he.chi.he.chi.te he.o.he.o.ta mo.i.mo.i.na ri.u.ri.u.ta

While all cases of disyllabic reduplication involve trisyllabic words, not all trisyllabic words have disyllabic reduplicants: there are such examples as i.vak.ta 'embrace' i. 'i.vak.ta and ma.ve.ta 'receive'-^ mav.ve.ta (Molina, Valenzuela, and Shaul 1999). Barragan and Haugen (2002) discuss the possibility that the Yaqui CV(C)V- reduplication occurs in a verb class of 'body movements and body functions', but this speculation would need to be tested more systematically. In terms of reconstruction, I leave as an open question whether or not this CV(C)V- reduplication would be better analyzed as full root reduplication as opposed to foot reduplication, since the marked heavy syllable reduplicant, with two moras, also constitutes a foot. 4 We will return to this issue when we discuss other cases of CV(C)V- reduplication in UtoAztecan.

2.4. Yaqui mora affixation The final kind of habitual reduplication in Yaqui might not be "reduplication" in a technical sense, but it does involve the repetition of material within a base word. This is the creation of a heavy syllable within the first syllable of the verb root which usually leads to consonant doubling, but also vowel-lengthening in words with medial consonant clusters (e.g. 7e):

Reduplicative allomorphy and language prehistory in Uto-Aztecan (7)

321

Yaqui habitual mora affixation (Molina et al. 1999) a. b. c. d. e.

b w a.ta.ni.a e.ta.po ho.vo.a ma.ve.ta yep.sa

'burn (food)' 'open up' 'get full' 'receive' 'arrive'

b w at.ta.ni.a et.ta.po hov.vo.a mav.ve.ta yeep.sa

Following Samek-Lodovici (1992), Haugen (2003) analyzes these as instances of mora affixation: the morpheme consists of a bare mora affix. 5 The phonological expression of habitual in these cases is related to the marked heavy syllable reduplication, and involves the enhancement of a phonological string with an additional mora, realized via the emergence of the unmarked. According to markedness theory, this extra mora could be realized crosslinguistically as gemination, vowel-lengthening, or epenthesis of an unmarked consonant (e.g. [?] or [h]), each of which we will see in various of the Uto-Aztecan languages to be discussed below.

2.5. Summary of Yaqui reduplication facts To sum up the facts that we have covered so far, Yaqui has four distinct patterns of reduplication. Three of these can be used to signal habitual action (among other possible meanings): a light syllable reduplicant, a disyllabic or full root reduplicant, and a bare-mora affix. These reduplication patterns are not phonologically predictable from the underlying root, therefore roots must belong to specific classes in order to determine which root goes with which reduplicant. This can be seen by the existence of nearminimal pairs, such as those in (8): (8)

a. b. c.

ivakta kinakta mäveta

'embrace' 'squint, grimace' 'receive'

i.'i.vak.ta ki.na.ki.nak.ta mav.ve.ta

In addition, Yaqui also has an additional (non-habitual) dupleme: "marked heavy syllable reduplication", which is a heavy syllable reduplicant with gemination from the base.

322 3.

Jason D. Haugen Reduplicative allomorphy across Uto-Aztecan

Each of the above reduplication patterns in Yaqui has cognates in other Uto-Aztecan languages. For each pattern of reduplication I will illustrate its cognate pattern in at least one SUA and one NUA language. The interesting thing about comparative Uto-Aztecan reduplication patterns is not merely that each of these languages has some form of reduplication, but that these languages often utilize multiple marked patterns of reduplication.

3.1. Light syllable reduplication From the available evidence it appears that syllabic reduplication, either of the form CV, CVV, or CVC, is the most common form of reduplication in the Uto-Aztecan languages. As a sample, syllabic reduplication is attested in at least Nahuatl (Tuggy 1979, Canger 1981), Cora (Casad 1984), Yaqui (Demers, Escalante, and Jelinek 1999), Mayo (Hagberg 1993), Guarijio (Miller 1996a), Tarahumara (Copeland 1992), Tohono O'odham (Fitzgerald 2000), Pima (Riggle 2001), Northern Tepehuan (Bascom 1982), Southeastern Tepehuan (Willett 1982), Hopi (Hendricks 1999), Serrano (K. Hill 2001), Gabrielino (Munro 1983), Northern Paiute (Snapp, Anderson, and Anderson 1982), Southern Paiute (Sapir 1930), and Comanche (Charney 1993). Although there may have been some diachronic development in particular languages which has led to a default heavy syllable reduplicant in those languages, because of the common relationship between a light syllable reduplicant and a "marked heavy syllable" variation on the light syllable reduplicant (where the "secondary" reduplication adds a variation on the "primary" light syllable reduplication by means of the addition of an extra mora), I reconstruct light syllable reduplication as the unmarked syllabic reduplication pattern in the pro to-language. This pattern is extremely common across the Uto-Aztecan family.

3.1.1. SUA: Mayo light syllable reduplication

(Hagberg

1993)

Mayo is Yaqui's closest linguistic relative, and the Spanish missionaries who first recorded them regarded them as dialects of the same language, Cähita. However, there are grammatical differences between them, including in the area of reduplication.

Reduplicative allomorphy and language prehistory in Uto-Aztecan

323

Mayo has cognates for both the light syllable reduplicant and the marked heavy syllable reduplicant of Yaqui. However, Hagberg (1993) reports that the semantic distinction between the two has been lost in all but the eldest generation of speakers. For younger speakers, there is free variation between the light and heavy syllable reduplication patterns in the expression of habitual action. In Spaelti's terms, what were once duplemes have merged into alloduples of a single dupleme. Campbell and Muntzell (1989) point out that one of the symptoms of language loss is the development of variability of this kind. Mayo is in more immediate jeopardy than Yaqui, but both are certainly endangered languages (see Moctezuma Zamarron 2001 for further discussion of the ethnographic context of language shift in the Yaqui and Mayo communities, and the differences between the two). Hagberg (1993) divides the Mayo lexicon into two classes: the accented class, with accent on the first vowel (9), and the unaccented class, with accent on the second vowel (10). These two classes differ with respect to reduplication: (9) a. b. c. d. (10) a. b. c. d.

Mayo accented words Stem REDl=q RED2=O l

Unattested

Gloss

yu.ke wom.te nok.wa no.ka

yiiy.yu.ke worn.worn.te nok.nok.wa non.no.ka

*yuk.yu.ke *wow.wom.te *non.nok.wa *n6k.no.ka

'rain' 'be frightened' 'known language' 'know a language'

Mayo unaccented words Stem RED 1=0, RED2=On

Unattested

Gloss

bwa.nä bwi.kä om.te no.ka

*bwab.bwa.na *bwib.bwi.ka *o\'om.te *non. no.ka

'cry' 'sing' 'hate' 'speak'

yu.yu.ke wo.wom.te no.nok.wa no.no.ka

bwa.bwä.na bwi.bwi.ka o.om.te no.no.ka

bwan.bwa.na bwik.bwi.ka om.om.te nok.no.ka

The contrast between the heavy syllable reduplicant (RED2) in (9d) non.no.ka and (lOd) nok.no.ka shows that the base for reduplication must be different in the two classes of words. Haugen (2004) discusses this difference in terms of an opacity in base-assignment: the right edge of the base is coterminous with the right edge of the accented syllable in the unreduplicated form.

324

Jason D. Haugen

Although the specifics of the phonology of Mayo reduplication differ somewhat from the phonology of reduplication in Yaqui, it is clear that the patterns here are cognate with those in Yaqui.

3.1.2. NUA: Gabrielino CVreduplication

(Munro 1983)

Gabrielino was a language of the Takic sub-branch of NUA. Examples of light syllable reduplication in Gabrielino are given in (ll): 6 (11)

a. b. c. d. e.

tokoo-r piino-r soo-t woose' camee-r

'woman' 'hummingbird' 'rattlesnake' 'dog' 'owl'

to-tooko-m pe-piino-r-am so-soo-t-om wo-woose'-am ca-caame-r-am

'women' 'hummingbirds' 'rattlesnakes' 'dogs' 'owls'

Because the reduplicant surfaces with a short vowel, even when the base has a long vowel, these data clearly illustrate that the reduplicant is a light syllable.

3.2.

Marked heavy syllable reduplication

Although there is evidence of marked heavy syllable reduplication in both the northern and southern languages, it is not clear that this pattern is still completely productive in most languages. As noted above, it is even apparently being lost by some speakers of Yaqui and Mayo. However, it is extremely productive in Tohono O'odham (formerly Papago), even to the extent that there are several different kinds of marked heavy syllable reduplication in this language.

3.2.1. SUA: Tohono O'odham There are several patterns of "marked heavy syllable reduplication" in Tohono O'odham, which are instances of multiple "duplemes". The first is the case of the so-called "marked plurals", which involve special phonological marking for a semantically-motivated class of noun plurals (Hill and Zepeda 1994). The usual pattern of plural marking in Tohono O'odham is the use of light syllable reduplication of the form CV-:

Reduplicative allomorphy and language prehistory in Uto-Aztecan (12)

325

Tohono O'odham plurals (Hill and Zepeda 1994) a. b. c. d. e.

bitokoi ce:kol cu:wl gogs mi:stol

bibtokoi ceckol cucu:wT gogogs mimstol

'Pinacate beetle' 'squirrel' 'jackrabbit' 'dog' 'cat'

There have been some recent theoretical discussions about whether the reduplicant is actually a prefix with syncope in the base, or whether it is an infix. Riggle (2001) has recently argued for the latter in the closely related language Pima (Akimel O'odham), while most scholars (e.g. Hill and Zepeda 1994, Fitzgerald 2000) have accepted the former for Tohono O'odham. Based on the typological markedness of infixing reduplication, the comparative Uto-Aztecan evidence, as well as the fact that syncope is independently attested in Tohono O'odham, I follow the prefixation-withsyncope analysis of Hill and Zepeda and Fitzgerald here. In the case of a special culturally-significant class of roots (less than 100 roots, according to Hill and Zepeda 1994), the plural is marked with a long vowel: (13)

Tohono O'odham long-vowel plurals (Hill and Zepeda 1994) singular plural a. b. c. d. e.

ban mad nem bahi §on

ba:ban ma:mad ne:nem ba:bhai §o:§on

'coyote' 'younger sister's child' 'liver' 'tail' 'trunk of a plant'

In addition to this pattern of marked heavy syllable reduplication, there are also cases of marked heavy syllables which trigger gemination, as with distributive nouns and verbs (14) and repetitive verbs (15). (14)

O'odham distributives (with gemination) (Fitzgerald 2003) singular plural distributive gloss a. b. c. d.

nowiu nahagio hodai ?a:g

nonowiu nanhagio hohodäi ?a?ag

nonnowiu nannhagio hohhodai ?a??ag

'ox' 'earring' 'rock, stone' 'a pair of animal horns'

326

Jason D. Haugen

(15)

O'odham repetitive verbs (with gemination) (Fitzgerald 2003) unitative repetitive gloss a. b. c. d. e.

giw hihim huhulgat kow jiia

giggiw hihhim huhhulgat kokkow jiinna

'hit something' 'laugh' 'menstruate' 'dig something out of the ground' 'look in a certain direction'

The approach to these data that I will advocate below is that these reduplicated forms are prosodically identical - they consist of a heavy syllable reduplicative affix, but their surface phonology is dependent upon the interaction of constraints. In O'odham, the usual heavy syllable reduplication pattern is with gemination. However, in order to make morphological distinctions (or to "enhance contrast", in the terminology of Urbanczyk this volume) between distributive nouns and "marked human" plural nouns, a different method of meeting the heavy syllable requirement is advanced: vowel lengthening. As I will illustrate in more detail in Section 5.1, this is easily explained by the relative ranking of markedness and faithfulness constraints.

3.2.2. SUA: Guarijio (Miller

1996a)

Reduplication in Guarijio verbs typically marks plural agreement, iteration, or duration. Guarijio nouns, like nouns in Tohono O'odham, have special morphology for human plurals; indeed, special marking on plural human nouns can actually be reconstructed for PUA (see Hill and Hill 2000). Plurals in Guarijio are normally not marked at all on nouns, but human nouns reduplicate, usually at least optionally with a heavy syllable. In the case of Guarijio, however, these marked plurals show up not with gemination or a long vowel, but with an epenthesized laryngeal (usually a glottal stop): (16)

Guarijio "marked heavy syllable" = "marked plural" (Miller 1996a") a. b. c. d. e.

ku.ci.ta ma.la.la se.pu.ri pa.mi.la no.la

ku?.ku.ci ma/ma?.ma.la.la se/se?.se.pu.ri pa?.pa.mi.la no/no?.no.la

'son, daughter' 'daughter' 'uncle, aunt' 'boss, govenor' 'son'

Reduplicative allomorphy and language prehistory in Uto-Aztecan

327

It is not implausible that these marked plurals are the remnants of what used to be a more complicated system, where non-human nouns were once reduplicated with a light syllable for the plural, but which reduplication was lost after verbal reduplication began to indicate plural agreement. Thus, I regard these heavy syllable reduplication patterns as cognate with the marked heavy syllable pattern seen elsewhere in Uto-Aztecan.

3.2.3. SUA: Nahuatl (Canger 1981) The various dialects of Nahuatl also utilize multiple patterns of reduplication, some of which are easily considered to be instantiations of "marked heavy syllable reduplication". Canger (1981) reviews data from across the Nahuatl dialects and convincingly argues for a reconstruction of at least three patterns of reduplication for Proto-Aztecan. The first is reduplication with a short vowel. The other two patterns are clearly instantiations of marked heavy syllable reduplication: CV?- reduplication for 'distributive' and CV:- reduplication for 'consecutive'. Canger shows that these patterns must have been productive in Proto-Aztecan, and the evidence surveyed in this paper suggests further that these patterns are the remnants of the marked heavy syllable reduplication pattern which was productive in ProtoUto-Aztecan itself. Of particular note here is the similarity of the morpheme for 'distributive' in Nahuatl (a reduplicant with a heavy syllable brought about via glottal stop insertion) and in Tohono O ' o d h a m (a reduplicant with a heavy syllable brought about via the gemination of the onset of the base). The correspondence of this morpheme in Nahuatl and Tohono O ' o d h a m strongly suggests a reconstruction of marked heavy syllable reduplication for 'distributive' in at least some intermediate branch within Uto-Aztecan, and the phonological difference between the exponence of this morpheme in Nahuatl and Tohono O ' o d h a m can be easily brought about via the reranking of t w o constraints (for further discussion see section 5.1).

3.2.4. NUA: Marked plurals in the NUA languages In Northern Uto-Aztecan, heavy syllable reduplicants also appear, in the context of plural nouns, in at least Comanche and Tümpisa Shoshone, both of which are within the N u m i c branch of N U A . However, in these cases the

328

Jason D. Haugen

process does not seem to be productive, and the heavy syllable reduplicants may be fossilized (examples cited in Hill and Hill 2000): Comanche number marking (Charnev 1993) a. b. c. d. e. (18)

nam.mi pi.a ta.ka ten.see pi.e.ti

'sister' 'big' 'relative' ' ten cents' 'elderly person'

na.na.na.mi pi.pi.a [pi.via] tah.ta.ka.nii teh.ten.see pih.pie.ti.nii

'sisters' 'big group' 'relatives' 'ten cents apiece' 'group of elders'

Tümpisa Shoshone number marking (Dayley 1989) a. b. c. d. e. f.

pa.pi

'older brother' pa.pi.am.mü pap.pa.pi.nan.ku pat.si 'older sister' pap.pat.si.am.mü pe.tü 'daughter' pep.pe.tüm.mü tangum.mü 'man' tat.tangung.ku tokkwapü 'aunt' tot.tok.kwapüammü tua 'son' tut.tu.am.mu

'older bro. (pi.)' 'older bro. (dl.)' 'older sister (pi)' 'daughter (pi)' 'man (dl)' 'aunt (pi.)' 'son (pi.)'

In Comanche, the heaviness of the syllable is indicated by means of laryngeal (h) insertion, this apparently being preferred even to fully copying a coda consonant from the base (e.g. ten.see -> teh.ten.see and not *ten.ten.see). In Tümpisa Shoshone the heaviness appears with gemination from the onset consonant to the coda of the reduplicant. While these examples of heavy syllable reduplicants might be fossils, they are at least suggestive that the pattern was once productive in the Numic languages.

3.3. Foot or full root reduplication The CV(C)V- reduplicant is the most nebulous of the reduplicative categories that we will explore, since "marked heavy syllable reduplication" is also technically comprised of a metrical foot. Thus, CV(C)V- reduplication might in fact better be reconstructed as full-root reduplication.

Reduplicative allomorphy and language prehistory in Uto-Aztecan

329

3.3.1. SUA: Mayo (Hagberg 1989) Hagberg (1989) provides several examples of CV(C)V- reduplication in Mayo. Most of these occur with adverbials (i.e. stems ending with the suffix —ti). The CV(C)V- reduplication indicates the repetition or distribution of the action indicated by the stem. (In this Spanish-based orthography j = h). (19)

Mavo CV(C)V- reduplication (Hagberg 1989) a. b. c. d. e.

ko.lop.ti po.roj.ti re.bek.ti ro.bak.ti ro.'i.ti

'suddenly' 'scattered' 'broken into pieces' 'rounded' 'limping'

ko.lo.ko.lop.ti po.ro.po.roj.ti re.be.re.bek.ti ro.ba.ro.bak.ti ro.'i. ro.'i.ti

3.3.2. NUA: Tübatulabal and PNU A Langacker (1977) presents Tübatulabal as the exemplar of CVCV- reduplication in Uto-Aztecan, including the following case in point: (20)

Tübatulabal (Voegelin 1935, cited by Langacker 1977) pisaa-bisaa'a-t RED-go-PRES

'He is going out repeatedly'

Langacker goes on to suggest that "distributive value would not be implausible as a reconstruction for this type [in PUA], especially if this is taken as including temporal distribution" (129). However, Heath (1978) reconstructs CVCV- reduplication (along with a suffix -?. . .ki) for 'iterative' in Proto-Northern-Uto-Aztecan, as opposed to CV- reduplication for Proto-Southern-Uto-Aztecan. Heath, following tradition in Uto-Aztecan linguistics, was assuming at that time that PUA roots consisted of CVCV- forms. However, recent work (especially various papers by Manaster Ramer) has begun to reconstruct word-medial consonant clusters for PUA. Manaster Ramer (1993) has also reconstructed some PUA forms with final consonants, thus being of the shape CVCVC. The issue of whether the correct reconstruction for the disyllabic pattern should be CV(C)V- or reduplication of the full root would perhaps be clarified if we found some language with a CVCVC- reduplication pattern, but I am not familiar with such a pattern in any Uto-Aztecan language.

330

Jason D. Haugen

Nonetheless, Langacker (1977), Bybee, Perkins, and Pagliuca (1994), and others assume that smaller reduplicants are usually reductions of larger reduplicants, and this seems to be a relatively safe supposition. The retention of the CV(C)V- reduplication pattern in both NUA and SUA means that PUA probably still had the CV(C)V-type reduplication simultaneously with the other types, although most daughter languages may have subsequently lost one or more of the patterns. The issue of grammaticization of smaller reduplicants from larger reduplicants will be taken up in section 5.2.

3.4.

Mora affixation

The final type of "reduplication" is mora affixation, which can manifest itself by means of morphological gemination, vowel-lengthening, or via epenthesis of an unmarked consonant such as glottal stop or [h],

3.4.1. SUA: Tepecano (Mason

1916)

Tepecano, now extinct, was a Tepiman language (i.e. a member of the same sub-branch which contains Tohono O'odham). According to Mason (1916), reduplication was used to mark the plural in the usual case. Examples (without accents) are given in (21): (21)

Tepecano light syllable reduplication ( = 'plural') a. b. c. d. e.

a.toc.kar ÖC

upp nov du:r

'seat' 'corn-field' 'skunk' 'hand' 'ant'

a.'a.toc.kar ö.'öc u.'upp no.nov du.du:r

However, Mason noted that "a second type of plural formation is found with disyllabic stems where the change ... occurs within the stem itself' (p. 330). This plural formation includes both gemination and glottal-insertion (and occasionally, both). There is no obvious semantic basis for the various classes.

Reduplicative allomorphy and language prehistory in Uto-Aztecan (22)

Tepecano morphological gemination ( = 'plural') a. b. c. d. e.

(23)

331

i:.pu:rr ko.ko:n hi.ku:rr ho.dai ia.puc.kar

'skirt' 'crow' 'peyote' 'stone' 'sweat-cloth'

i:p.purr kok.kon hik.kur hod.dai iap.puc.kar

Tepecano glottal-insertion ( = 'plural') a. b. c. d. e.

ta.tak go.goc i.ma.i du.du:r asa:k

'nerve' 'dog' 'squash' 'jaguar' 'net'

ta't.tak go'.goc i'.ma.i du'.dur a'.sak

(Note the homophonous dudu:r 'ant (pi.)' and dudu.r 'jaguar (sg.)'·) Regarding his transcription of the data, Mason noted that "[wjhile it is a delicate task to analyze correctly the phonetic characteristics of these plurals, and some of the above are probably not accurately expressed, yet the general process of secondary plural formation seems to be... that a medial stop is lengthened or a glottal stop introduced..." (p. 331, emphasis added). Two important things about these data are clear. First, there was a process of mora affixation in this language, the phonetic realization of which was apparently ambiguous between consonant doubling and laryngealinsertion. Second, the system of plural-marking in this Tepiman language was different than that of its close relatives (e.g. Tohono O'odham), and the loss of this language may have forever obscured the historical development of reduplication and mora affixation within Tepiman, and ultimately perhaps larger groupings within Uto-Aztecan as well.

3.4.2. NUA: The Numic languages (McLaughlin 2001) The Numic languages are a relatively closely knit group of languages in NUA. McLaughlin (2001) provides evidence that all of the Numic languages share the morphological process of mora affixation. 7 In these languages, this morphological process marks various aspectual distinctions, and sometimes agreement. McLaughlin reconstructs the proto-Numic pat-

332

Jason D. Haugen

tern to be gemination marking momentaneous aspect, singular (ergative) agreement, and "move while doing" as opposed to "move to do". This gemination appears on either the verb stem or on motion suffixes. Examples from Central Numic are given in (24): (24)

Central Numic Panamint Shoshone Comanche

a. b. c. d. e.

kati yake kati yakai kati

'sit' 'cry' 'sit' 'cry' 'sit'

kat.ti yak.ke kat.ti yak.kai katti

'sit (dur.)' 'cry (dur.)' 'sit (dur.)' 'cry (dur.)' 'sit (dur.)'

These examples from Central Numic are clear instances of morphological gemination, one instantiation of mora affixation. The Western Numic examples below require a bit more explanation. The Northern Paiute examples given by McLaughlin show devoicing of a segment for the durative. The appearance of a voiceless stop here indicates underlying gemination, otherwise in this language we would expect spirantization of the intervocalic voiceless consonants (Snapp et al. 1982). (25)

Western Numic Northern Paiute a. yaga 'cry' b. zoba 'gather'

yaka zopa

'cry (dur.)' 'gather (dur.)'

Besides McLaughlin's list of Northern Paiute duratives, Snapp et al. (1982) also include the following: mia mi'a 'go'; nrni nimmi 'move about/walk/go'; tiikwi tiikwi'i 'tell'; himma -> hi 'ma 'carry (pi.)'; and sumaya sumaya 'remember'. Thus, in Northern Paiute, in addition to gemination, we can also see epenthesis of glottal stop and epenthesis of a copy of the final vowel of the word for the durative morpheme. It is also clear from the discussion of Snapp et al. that these reflexes of mora insertion are independent from syllabic reduplication used for number agreement, which is prefixal for 'dual' (e.g. mia mi.mia 'go (dual)'), and, occasionally, suffixal for 'plural' (e.g. mia mia. 'a 'go (pi.)'). McLaughlin gives examples from Mono which apparently show the consonant-insertion variation of mora affixation: the mora is realized as the unmarked consonant [h].

Reduplicative allomorphy and language prehistory in Uto-Aztecan (26)

Mono

a. b. c. d. e.

cahqoti wihtapo wihtahpo'i kwaca kwahca'i

333

'break several brittle objects' 'strike several blows' 'strike one blow' 'fall' 'fall a short distance'

The data from Southern Numic are less clear, but the doubling of consonants within the stem does suggest the process of mora affixation: (27)

Southern Numic Southern Paiute a. b.

iya-paqa uqwi

'fear' 'smell'

a', b'.

yi'i-ppaqqa 'be afraid several times' u'uqqwi 'sniff around'

In addition to these data from Numic, Heath (1978) reconstructs "hardening" (i.e. glottal insertion) for punctual aspect in PNUA: CVCV-ki 'durative' -> CV'CV-ki 'punctual' (e.g. Mono qwaca 'fall' -> qwaca-ca-hi (durative) q a'ca-'i- (iterative)). Heath also noted that Southern Paiute and Luiseno (an N U A language from the Takic sub-group) retain this distinction with a vowel-length alternation (although no data is given). All of these examples indicate the presence of a bare-mora affix in NUA.

3.5. An interesting additional Uto-Aztecan pattern: #V- reduplication One additional pattern of reduplication, which is marked crosslinguistically, also occurs in Uto-Aztecan. This is the pattern where the reduplicant only copies the first vowel of the base (cf. Steriade 1988, who points out that the reduplicant typically has a less marked structure than the base, but here, where the reduplicant has no onset, it is more marked than the base). This occurs in Tübatulabal (NUA) and Tarahumara (SUA). Examples from Tübatulabal are given in (28):

334

Jason D. Haugen

(28) ( Ν υ Α Ί Tübatulabal "Initial Reduplication" (Voegelin 1935) Regular formation telic telic a. b. c. d.

elat'lktanapa:abi-

e.'ela "i.tiik an.dana a:.ba:ab'i

Reversed formation atelic telic e. f. g· h.

a.naqa.nabi.cilu:b ö.tölo:h

nai) nap cilu:p tolo:h

gloss 'to 'to 'to 'to

jump' eat' get down' be tired'

gloss 'to 'to 'to 'to

cry' throw' split wood' groan'

Particularly noteworthy here is that the function of this Tübatulabal reduplication pattern is to make telic verbs atelic and atelic verbs telic. Alderete et al. (1999) discuss Tübatulabal reduplication and show that these forms actually have a glottal stop in onset position, as well as homorganic nasal codas in some forms, both of which arise via emergence-of-theunmarked effects brought about through high-ranking markedness constraints and low-ranking MAX-BR. A very similar pattern exists in the SUA language Tarahumara, which is closely related to Guarijio. According to Copeland (1992), Tarahumara reduplication indicates frequency, iteration, degree of intensity, habitual/tendency, plurality or distributivity. Copeland notes that "the major pattern of reduplication involves the stressed copying of the first CV sequence" (p. 315): (29)

a. b. c.

yo-ma hi'i-ma wi-'ma

'to be angry' 'to run' 'to bind with rope'

yo-yo-ma hu-hu-ma wi-'wi-ma

'to be very angry' 'to race' 'to lasso'

However, reduplication usually involves concomitant intensification or weakening of contiguous (i.e. base) syllables, and the reduplicant can appear without a copy of the onset consonant of the base:

Reduplicative allomorphy and language prehistory in Uto-Aztecan (30) a. ri-pi-ma 'to stay' i-'ti-bi-ma 'to stay frequently' *ri-'ti-bi-ma

335

b. bo-ci-ma 'to fill up on food' ο-'ρό-ji-ma 'to fill up repeatedly' *bo-'p0-ji-ma

Copeland attributes this onset loss to the historical factor of stress-shift to the right—stress in Uto-Aztecan is usually on the first or second syllable, but in Tarahumaran (that sub-group of SUA which contains the dialects of Tarahumara and Guarijio) it is usually on the third syllable. Copeland also argues that this reduced reduplicant has in some cases been grammaticized (i.e. phonologically eroded) into a morpheme which has lost the phonological reduplication and has been reinterpreted as a simple augmenting vowel with the rough approximation of the semantics of the original reduplication: li-l. Copeland says that li-l is the most widespread "because of the prevalence of reduplicated forms containing an original ///" (p. 318), but it seems plausible to me that, synchronically, [/-] is just a default epenthetic vowel. In sum, for Tarahumara Copeland reconstructs CV- reduplication with onset loss after right-ward stress shift. 8 A similar historical analysis might also be appropriate for Ttibatulabal. Although this interesting pattern of V-initial reduplication occurs in both an N U A and an SUA language, I think that it is plausible that this was an independent development in each of these two languages, and I do not reconstruct this pattern for PUA.

4.

Reduplication and diachrony in Uto-Aztecan

4.1. Reconstructing reduplication patterns for PUA In the most well-articulated formulation of Uto-Aztecan comparative grammar to date, Langacker (1977:128) notes that "virtually every UA language displays verbal reduplication of some kind, and in some cases a variety of patterns", but he was unable at that time to suggest a definitive reconstruction for the Proto-Uto-Aztecan (PUA) reduplication pattern(s). He did illustrate a variety of reduplication patterns across the Uto-Aztecan family (without referring to prosody): CVCV-, CVC-, CV?-, CV:-, CV-, and he suggested that the smaller reduplicants are probably reduced from the larger ones. Langacker also noted the equivalent of what I have been referring to

336

Jason D. Haugen

as mora affixation: consonant gemination (or non-lenition, as in Northern Paiute), vowel-lengthening, and glottalization, as well as accent shift. Heath (1978) made three specific reconstructions for Proto-NorthernUto-Aztecan reduplication: (31)

Heath's reconstructions for PNU A (1978) i. ii. iii.

CVCV-: CV-: -?-:

'iterative' 'distributive' 'punctual'

The first of these is obviously related to what I have been calling disyllabic or full root reduplication; the second appears to be light syllable reduplication. However, with the function of 'distributive' being performed by marked heavy syllable reduplication in multiple languages in SUA, exploring this possibility for N U A might alter this picture a bit. The final morpheme, presented by Heath as a glottal stop infix, is actually a bare mora affix, which surfaces in some Numic languages (and Guarijio) as a glottal stop or [h], but in other languages can also be realized as gemination or a long vowel. My specific proposal for the reconstruction of the Proto-Uto-Aztecan reduplication patterns is the following (leaving the original meanings to future research, pending more complete paradigms from more languages). PUA had at least four reduplicative morphemes (i.e. duplemes): (32)

Proto-Uto-Aztecan reduplication patterns i. light syllable reduplication ii. "marked heavy syllable" reduplication iii. full root (or disyllabic) reduplication iv. mora affixation

The second and fourth patterns are relatively marked cross-linguistically. Each of the four patterns occurs in multiple languages of each of the major branches of the family, and can therefore be reconstructed for the protolanguage. The next step in this research will be to try to elucidate the functions of each of these patterns in all of the Uto-Aztecan languages, in order reconstruct the original patterns in Proto-Uto-Aztecan and subsequently to retrace the semantic evolution of each of these morphemes in the daughter languages.

Reduplicative allomorphy and language prehistory in Uto-Aztecan

337

4.2. Searching for larger connections for Uto-Aztecan There have been various claims in the literature relating Uto-Aztecan to other North American language families, and there have been a number of (often wildly divergent) proposals and hypotheses, of which many have been based on lexical evidence alone. Some of the more prominent proposals include linking Uto-Aztecan with Tanoan (Whorf and Trager 1937); with Kiowa, Tanoan, and Zuni (Sapir 1929); with Kiowa-Tanoan and Keresan (Davis 1979); with Kiowa-Tanoan and Oto-Manguean ( ~ "Central Amerind") (Greenberg 1987); with Penutian, Mayan, Kiowa-Tanoan and Totonac ( ~ "Macro-Penutian") (Whorf 1935); with Penutian, Wakashan, Salishan, and Algonkian-Ritwan (Liedtke 1996); and with Mixe-Zoquean (Wichmann 1999). Matching two, three, or even all four of these reconstructed reduplicative morphemes (or perhaps even one of the typologically marked ones) would strengthen the claim of any proposed genetic relationships between Uto-Aztecan and other language families.

4.3. Against (purely) areal influence on the sharedness of these patterns Sherzer (1976) examines the various "linguistic areas" of the United States (for Sherzer, these correlated with cultural areas) in which Uto-Aztecan is located, and although Uto-Aztecan is not limited to these areas, he came up with the following observations. In the Southern California area (one of three areas in California), Uto-Aztecan and Yuman both have verbal and nominal reduplication. Sherzer surmised that reduplication is a family trait of Uto-Aztecan, and perhaps (verbal) reduplication diffused to Yuman languages, and (nominal) reduplication diffused to other southern California languages as well. Sherzer comments on reduplication in the Southwest area with the following remark: "It is interesting to note that this very common North American Indian linguistic trait is rather rare in the Southwest" (144), although reduplication is a familial trait of Uto-Aztecan. In the Southwest, Great Basin and Southern California, verbal reduplication is a regional areal trait, and a family trait of Uto-Aztecan. The Great Basin area consists of Washo [Hokan] and many of the Numic languages. Sherzer hypothesizes that reduplication may have been diffused from the Numic languages into Washo. In the Plains, the only Uto-Aztecan language is Comanche, but

338

Jason D. Haugen

other language families represented include Siouan and Tonkawan. Sherzer says that reduplication has been retained by languages in which it is a family trait (Uto-Aztecan, Siouan), but has not developed in other languages of the Plains area (Tonkawan). As far as linguistic areas south of the United States are concerned, Suärez (1983) notes that reduplication occurs in Uto-Aztecan, Mayan, Mixe-Zoquean, and does not occur in Tequistlatec, Totonac, or Otomanguean. There is very little comparative data presented by Suarez, but the former would be interesting places to look for further information to see if there can be links between Proto-Uto-Aztecan and some other macro-level language grouping. Given the extreme geographical range of the Uto-Aztecan family, however, it is highly unlikely that contact brought each of these patterns into each of the major branches after the PUA split. Regarding the establishment of genetic relationships, Bynon (1981) remarks that "the best proof that a language belongs to a certain family is that in it relics of former grammars...are preserved as synchronic irregularities" (70, cited in Liedtke 1996). On this point, reflect back to the fossilized remnants of marked heavy syllable reduplication in the plurals of Comanche and Tümpisa Shoshone, which, I claim, show such an irregularity. There is no doubt that Numic is related to Southern Uto-Aztecan in a genetic relationship (as established by Sapir 1913), therefore such an irregularity must have come from a formerly productive process, i.e. marked heavy syllable reduplication in Proto-Uto-Aztecan (and not PNUA because the process is still very active in the southern languages). If it turns out that there is linguistic evidence for very early contact among the PUA community and some other linguistic group, then this information might be able to be used as additional evidence in the current debate regarding the origins of the PUA community: i.e. the Northern versus Southern homeland hypotheses (the former being the traditional view, represented by such scholars as Fowler 1983, and the latter being a recent proposal by Hill 2001). If other families could be shown to have similar patterns of reduplicative allomorphy then there would be a strong indication of very old contact and/or descent from a common ancestor among the PUA community and the other proto-community. A recent promising possibility is the relationship of Uto-Aztecan to Mixe-Zoquean, for which Wichmann (1999) gives a number of potential cognate sets. Although Wichmann's (1995) book on reconstructing Mixe-Zoquean phonology does not delve into reduplication, he has informed me (personal communication)

Reduplicative allomorphy and language prehistory in Uto-Aztecan

339

that reduplication in Mixe-Zoquean is exclusively suffixal. Thus, evidence from the comparative reduplication patterns discussed here will probably not provide us with a convincing link between these two families.

5.

Reduplication and diachrony more generally

In addition to the relevance of the Uto-Aztecan reduplication patterns for comparative Uto-Aztecan linguistics, the data surveyed in this paper are also relevant to theoretical approaches to reduplication, and to issues in the diachrony of reduplication more generally.

5.1. The interaction of constraints and ranking in reduplication First, the divergent realizations of the reduplicative allomorphs in different Uto-Aztecan languages support an approach to reduplication which refers to the relative ranking of markedness and faithfulness constraints. For instance, "marked heavy syllable" reduplication is variably realized by consonant gemination (e.g. the Tohono O'odham distributive), vowellengthening (e.g. the Tohono O'odham marked plurals), or epenthesis of an unmarked consonant (e.g. Guarijio plural human nouns). This is what we would expect from an approach to reduplication based on constraintranking where an additional mora is added to a phonological string and is realized according to independently-attested markedness constraints. It would be difficult to attain the cross-linguistic generalizations from an approach which ignores prosody and insists necessarily upon direct mapping between specific consonants and vowels in different languages. Three constraints are enough to do the requisite job: (33)

a. b. c.

DEP-IO: *LONG-V: *LONG-C:

No epenthesis. No long vowels. No gemination.

With these three constraints we can account for the full typology of marked heavy syllable reduplication found in Uto-Aztecan:

340

Jason D. Haugen

(34)

Markedness Typology in Uto-Aztecan "Marked Heavy Syllable Reduplication"

Constraint Ranking DEP-IO, »LONG-V *LONG-C

»

DEP-IO, *LONG-C, *LONG-V

»

*LONG-V, *LONG-C DEP-IO

»

Result

Example Language

gemination

Yaqui, Central Numic

vowel-lengthening

O'odham ("marked plurals"), Nahuatl

epenthesis ([?] or [h])

Guarijio, Nahuatl, Mono

According to this view, the constraint-ranking will determine what surfaces in the environment of mora augmentation. That is, what we observe is the emergence of the candidate that violates the lowest constraint. In some cases (e.g. Tohono O ' o d h a m and Nahuatl) there will also need to be a constraint to derive distinct morphemes. In these cases, the second lowest ranked constraint will determine what output form surfaces. Thus, what we see is the emergence of the least marked, since what is unmarked for one morpheme may be marked relative to other morphemes in the language.

5.2. Implications of the Uto-Aztecan data for Grammaticization Theory Bybee, Perkins and Pagliuca (1994) discuss the diachronic development of reduplication from the perspective of grammaticization theory, which holds that grammatical (i.e. functional) morphemes develop from lexical morphemes through historical paths brought on by semantic and phonological erosion. They discuss the potential problem that reduplicative morphemes pose for such a theory: namely, that they challenge "the principle that all grams develop from a fuller lexical source, since it is not possible to trace a reduplicative gram back to a single word or even a specific phrase" (p. 166). However, Bybee et al. suggest that their theory can in fact be applied to reduplicative morphemes, if we consider the "fullest, most explicit form of reduplication, total reduplication, to be the originating point for all reduplications, with the various types of partial reduplication as reductions and thus later developments from this fullest f o r m " (p. 166). Bybee et al. illustrate this perspective with data from languages, such as Trukese, where there are multiple patterns of reduplication in the synchronic state of the language. In Trukese, there are three distinct types of

Reduplicative allomorphy and language prehistory in Uto-Aztecan

341

(prefixal) reduplication (i.e. three different "duplemes"): total reduplication (with the meaning of iterative or continuative), syllable reduplication (with the meaning of habitual), and initial consonant doubling (which acts as an intransitivizer). The central crux of Bybee et al.'s argument is that the fuller reduplicant shape must have been the oldest, with the smaller and smaller reduplicants being later developments along a semantically plausible path of development. They state that "the modern situation in Trukese suggests that reduplication can grammaticize more than once in the history of a language, and that forms produced by successive waves of grammaticization can co-exist, although the form and meaning of each one identifies its age" (p. 173). We saw above that Yaqui has three reduplicative allomorphs serving a single semantic function - habitual action. All three of these reduplicants are similar in form to the semantically variant duplemes of Trukese. However, in Yaqui no appeal can be made to the semantics to determine the age of the various reduplicants. Although it certainly seems plausible that more full reduplicants may eventually reduce to smaller ones, perhaps along the semantic lines proposed by Bybee et al. for Trukese, the Yaqui data show clearly that it is not the case that the development of the reduplicants is necessarily evident at every stage of a language. The Yaqui data alone challenge the major conclusion that Bybee et al. draw from their stratified probability sample: namely, that their study "does show a clear association of meaning with form which reflects the diachronic development of reduplicative grams" (p. 173). The Yaqui allomorphs of reduplication, in the synchronic state of the language, show anything but a clear association between meaning and form reflecting the diachronic development of the reduplicative allomorphs. 9 However, these variable allomorphs in Yaqui did point to the need for the further investigation of the historical sources of reduplicative allomorphs within the larger context of the Uto-Aztecan language family. It is hoped that future research will be able to add more comprehensive data from other Uto-Aztecan languages in order to reconstruct the system of reduplication in Proto-Uto-Aztecan, from which the paths of grammatical change could then possibly be deduced. Such a project, with an elaboration of phonological and semantic changes and shared retentions, might also contribute to the ongoing debate regarding sub-branching within UtoAztecan as well.

342

Jason D. Haugen

5.3. Morphemic status of the reduplicant Finally, from the framework of Distributed Morphology (Halle and Marantz 1993, Harley and Noyer 1999) Raimy (2000) presents a piece-based theory of reduplication where the reduplicant surfaces via a shift in linear precedence in the underlying representation brought about by a re-write rule triggered by a null morpheme. I think it is plausible that the stability of the reduplicant forms across Uto-Aztecan shows that the reduplicant itself should be regarded as a morphemic piece, consistent with the grammaticization approach of Bybee et al. (1994) and Langacker (1977), among oth-

6.

Conclusion

In this paper we have surveyed evidence from across the Uto-Aztecan language family that suggests that Proto-Uto-Aztecan must have had a rich system of reduplicative morphemes, consisting of at least a light syllable, a heavy syllable with some kind of marking (gemination, laryngeal-insertion, or vowel lengthening), full root or CV(C)V- reduplication, and mora affixation. Pending the collection of more complete data from more of PUA's daughter languages, it may be possible to illustrate the historical trajectory of these morphemes from the proto-language to the present, and if so it might also be possible to use the shared retention or innovation of these morphemes to argue for specific sub-branchings within the Uto-Aztecan family.

Notes 1.

2.

I would like to thank Luis Barragan, Colleen Fitzgerald, Larry Hagberg, Mike Hammond, Cathy Hicks Kennard, Jane Hill, Bob Kennedy, Peter Norquest, Adam Ussishkin, and the Phonology Reading Group and the Friends of UtoAztecan at the University of Arizona for helpful discussion of earlier versions of this paper. Any remaining errors of fact or interpretation are mine alone. Besides these two possible innovations in SUA, Silver and Miller also mention a law of PUA **i -> PSUA *e. However, the Tepiman languages retain the /i/ vowel, therefore this is not a candidate for a legitimate SUA sound law.

Reduplicative 3. 4.

5.

6.

7.

8.

9.

allomorphy and language prehistory in Uto-Aztecan

343

The Yaqui language is also known as Hiaki and Yoeme. Historically speaking, the 'root' in the Yaqui disyllabic cases would be the CVCV- preceding the /h-ta/ or /k-ta/ endings, which may have some historical morphological purpose: the Ikl (and possibly /h/) perhaps being the remnants of the Uto-Aztecan "k-class", and the -(t)a and -(t)e being the remnants of morphological markers of transitivity (see Jelinek and Escalante 2000 for further discussion). Davis (2001) discusses similar cases of cross-linguistic mora-addition under the rubric of "mora augmentation," and he makes a distinction between prosodic and morphological mora augmentation. The examples discussed here are of the morphological type. Many (if not most) Uto-Aztecan languages utilize reduplication for nouns as well as verbs. In most cases here I have considered verbal reduplication, but since reduplication typically marks plurality of events or entities the noun/verb distinction is not relevant to our discussion, and at least some form of reduplication can be reconstructed for both nouns and verbs in PUA. The morphological effects of aspectual gemination are independent of phonological gemination brought about by the "final features" of certain Numic stems. For a discussion of this aspect of Numic phonology, with reference to Shoshone, see Miller (1996b: 694-697). Copeland's analysis of initial consonant loss in Tarahumara reduplication is supported by the observation that this language has also lost the initial onset (and sometimes first CV- sequence) of many Uto-Aztecan roots as well (see Miller 1985). Bob Kennedy (personal communication) has pointed out that Goodenough and Sugita (1980) also give examples of overlaps between the forms and functions of the reduplicants in Trukese itself.

References Alderete, John, Jill Beckman, Laura Benua, Amalia Gnanadesikan, John McCarthy and Suzanne Urbanczyk 1999 Reduplication with fixed segmentism. Linguistic Inquiry 30.3: 3 2 7 364. Barragan, Luis M. and Jason D. Haugen 2002 Reduplication in the Sonoran languages of Uto-Aztecan. In Sexto Encuentro Internacional de Lingüistica en el Noroeste, (Memorias Tomo 2), Rosa Maria Ortiz Ciscomani and Zarina Estrada Fernandez (eds.), 53-76. Hermosillo, Sonora: Editorial UniSon.

344

Jason D. Haugen

Bascom, Burton 1982 Northern Tepehuan. In R. Langacker, ed. Studies in Uto-Aztecan Grammar, vol. 3, Uto-Aztecan Grammatical Sketches. Arlington, TX: Summer Institute of Linguistics and The University of Texas at Arlington. Bybee, J. R. Perkins and W. Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Bynon, Theodora 1981 Historische Linguistik. München: Beck. Campbell, Lyle 1997 American Indian languages: The historical linguistics of Native America. Oxford: Oxford University Press. Campbell, Lyle and Martha C. Muntzell 1989 The structural consequences of language death. In Investigating Obsolescence: Studies in Language Contraction and Death, Nancy C. Dorian (ed.), 181-196. Cambridge: Cambridge University Press. Canger, Una 1981 Reduplication in Nahuatl in Dialectal and Historical Perspective. In Nahuatl Studies in Memory of Fernando Horcasitas, Francis Karttunen (ed.), 29-54. Austin, TX: Texas Linguistic Forum 18. Casad, Eugene H. 1984 Cora. In Studies in Uto-Aztecan Grammar, vol. 4: Southern UtoAztecan Grammatical Sketches, Ronald W. Langacker (ed.), 151459. Dallas, TX: Summer Institute of Linguistics. Charney, Jean Ormsby 1993 A grammar of Comanche. Lincoln, NE: University of Nebraska Press. Copeland, James 1992 Tarahumara reduplication: the grammaticization of iconic intensification. 19'h LACUS Forum: 313-335. Cortina-Borja, M. and L. Valinas 1989 Some Remarks on Uto-Aztecan Classification. International Journal of American Linguistics 55.2: 214-239. Davis, Irvine 1979 The Kiowa-Tanoan, Keresan, and Zuni languages. In The languages of Native America: Historical and comparative assessment, Lyle Campbell and Marianne Mithun (eds.), 390-443. Austin: University of Texas Press. Davis, Stuart 2001 Prosodic versus morphological mora augmentation. Paper presented at the High Desert Linguistics Society, 2001.

Reduplicative

allomorphy and language prehistory in Uto-Aztecan

345

Demers, Richard Α., Fernando Escalante and Eloise Jelinek 1999 Prominence in Yaqui words. International Journal of American Linguistics 65.1: 40-55. Dixon, R. M. W. 1997 The rise and fall of languages. Cambridge: Cambridge University Press. Escalante, Fernando 1985 A preliminary view of the structure of Yaqui. MA thesis, Department of Linguistics, University of Arizona. Fitzgerald, Colleen M. 2000 Vowel hiatus and faithfulness in Tohono O'odham reduplication. Linguistic Inquiry 31.4: 713-722. 2003 How prosodically consistent is Tohono O'odham? In Studies in UtoAztecan, Luis M. Barragan and Jason D. Haugen (eds.), 55-74. Cambridge, MA: MIT Working Papers on Endangered and Less Familiar Languages 5. Fowler, Catherine S. 1983 Some lexical clues to Uto-Aztecan prehistory. International Journal of American Linguistics 49.3: 224-257. Goodenough, Ward H. and Hiroshi Sugita 1980 Trukese-English dictionary. Memoirs of the American Philosophical Society v. 141. Philadelphia: American Philosophical Society. Greenberg, Joseph H. 1987 Language in the Americas. Stanford: Stanford University Press. Hagberg, Larry 1989 Reduplication in Mayo. Ms, University of Arizona and Summer Institute of Linguistics. 1993 An autosegmental theory of stress. Ph.D. dissertation, Department of Linguistics, University of Arizona. Halle, Morris and Alec Marantz 1993 Distributed morphology and the pieces of inflection. In The view from building 20: Essays in linguistics in honor of Sylvain Bromberger, Kenneth Hale and Samuel Jay Keyser (eds.), 111-176. Cambridge: MIT Press. Harley, Heidi and Maria Amarillas 2003 Reduplication multiplication in Yaqui: Meaning χ form. In Studies in Uto-Aztecan, Luis M. Barragan and Jason D. Haugen (eds.), 105— 139. Cambridge, MA: MIT Working Papers on Endangered and Less Familiar Languages 5. Harley, Heidi and Rolf Noyer 1999 State-of-the-article: Distributed Morphology. Glot International 4.4: 3-9.

346

.Jason D. Haugen

Haugen, Jason D. 2003 Allomorphy in Yaqui reduplication. In Studies in Uto-Aztecan, Luis M. Barragan and Jason D. Haugen (eds.), 75-103. Cambridge, MA: MIT Working Papers on Endangered and Less Familiar Languages 5. 2004 An opacity problem in Mayo reduplication. Kansas Working Papers in Linguistics 27: 139-156. Heath, Jeffrey 1977 Uto-Aztecan morphophonemics. International Journal of American Linguistics 43: 27-36. 1978 Uto-Aztecan *na-class verbs. International Journal of American Linguistics 44: 211-236. Hendricks, Sean Q. 1999 Hopi nominal reduplication: Prosodic non-uniformity across a paradigm. Proceedings ofWECOL 1998. Hill, Jane H. 2001 Proto-Uto-Aztecan: A community of cultivators in Central Mexico? American Anthropologist 103.4: 913-934. Hill, Jane H. and Kenneth C. Hill 2000 Marked and unmarked plural nouns in Uto-Aztecan. In Uto-Aztecan: Structural, temporal, and geographical perspectives, Papers in memory of Wick R. Miller by the Friends of Uto-Aztecan, Eugene H. Casad and Thomas L. Willett (eds.), 241-275. Hermosillo, Son: Universidad de Sonora. Hill, Jane H. and Ofelia Zepeda 1994 Tohono O'odham (Papago) plurals. In Estudios de lingüisticay sociolingiüstica, Gerardo Lopez Cruz and Jose Luis Moctezuma Zamarron (eds.), 13- 69. Hermosillo: Universidad de Sonora. Hill, Kenneth C. 2001 A typological sketch of Serrano. Presentation to the Uto-Aztecan Workshop, University of Arizona, 2001. Jelinek, Eloise and Fernando Escalante 2000 Unaccusative and unergative verbs in Yaqui. In Uto-Aztecan: Structural, temporal, and geographical perspectives, Papers in honor of Wick R. Miller by the Friends of Uto-Aztecan, Eugene H. Casad and Thomas L. Willet (eds.), 171-182. Hermosillo: Universidad de Sonora. Langacker, Ronald W. 1977 Studies in Uto-Aztecan grammar, vol. 1: An overview of Uto-Aztecan grammar. Arlington, TX: Summer Institute of Linguistics and the University of Texas at Arlington.

Reduplicative

allomorphy and language prehistory in Uto-Aztecan

347

Liedtke, Stefan 1996 The languages of the 'First Nations': Comparison of Native American languages from an ethnolinguistic perspective. Munich: Lincom Handbooks in Linguistics 01. Manaster Ramer, Alexis 1992 A Northern Uto-Aztecan sound law: *-c-y-. International Journal of American Linguistics 58.3: 251—68. 1993 Blood, tears, and murder: The evidence for Proto-Uto-Aztecan syllable-final consonants. In Historical Linguistics 1991: Papers from the 10,h International Conference on Historical Linguistics, Jaap van Marie (ed.), 199-209. Amsterdam and Philadelphia: John Benjamins. Marantz, Alec 1982 Re reduplication. Linguistic Inquiry 13.3: 435^182. Martinez Fabian, Constantino 1995 Reduplication in Yaqui and Optimality Theory. Ms, Universidad de Sonora and the University of Arizona. Mason, J. Alden 1916 Tepecano, A Piman language of Western Mexico. Annals of the New York Academy of Sciences, vol. XXV, 309-416. McCarthy, John and Alan Prince 1986 Prosodic morphology. Ms, University of Massachusetts-Amherst and Rutgers University. McLaughlin, John 2001 Momentaneous aspect and durative aspect in Numic languages. Paper presented at the Friends of Uto-Aztecan Conference, Santa Barbara, CA. 2001. Miller, Wick R. 1984 Classification of the Uto-Aztecan languages based on lexical evidence. International Journal of American Linguistics 50.1: 1-24. 1985 Lionnet's article on the "intensive" in Tarahumara. International Journal of American Linguistics 51.4: 502-504. 1996a Guarijlo: Grammatica, textos y vocabulario. Mexico City: Universidad Nacional Autonoma de Mexico. 1996b Sketch of Shoshone, A Uto-Aztecan language. In Handbook of North American Indians, vol. 17: Languages, Ives Goddard (ed.), 693-720. Washington D.C.: Smithsonian Institution Press. Moctezuma-Zamarron, Jose Luis 1998 Yaqui-Mayo language shift. PhD dissertation, Department of Anthropology, University of Arizona. Molina, Felipe, Herminia Valenzuela, and David L. Shaul 1999 Yoeme-English English-Yoeme standard dictionary: A language of the Yaqui tribe in the American southwest and northern Mexico, with a comprehensive grammar of Yoeme language. New York: Hippocrene Books.

348

Jason D. Haugen

Moravcsik, Edith 1978 Reduplicative constructions. In Universals of human language, vol. 3: Word structure, Joseph H. Greenberg (ed.), 297-334. Stanford: Stanford University Press. Munro, Pamela 1983 Selected studies in Uto-Aztecan phonology. International Journal of American Linguistics 49.3: 277-298. Raimy, Eric 2000 The phonology and morphology of reduplication. Berlin/New York: Mouton de Gruyter. Rehg, Kenneth L. 1981 Ponapean reference grammar. Honolulu: University Press of Hawaii. Riggle, Jason 2001 Phonotactics and reduplication in Pima. Paper presented at the 6th Southwest Workshop on Optimality Theory (SWOT 6). University of Southern California. Samek-Lodovici, Vieri 1992 A unified analysis of crosslinguistic morphological gemination. Proceedings of CONSOLE-1. Utrecht, The Netherlands [ROA-1491096], Sapir, Edward 1913 Southern Paiute and Nahuatl, a study in Uto-Aztecan. Journal de la Societe des Americanistes de Paris 10: 3 7 9 ^ 2 5 . 1929 Central and North American languages. Encyclopedia Britannica, 14th ed., London and New York: Encyclopedia Britannica Co. 1930 Southern Paiute: A Shoshonean language. In The collected works of Edward Sapir X, William Bright (ed.), 17-314. New York: Mouton de Gruyter. Sherzer, Joel 1976 An areal-typological study of American Indian languages north of Mexico. Amsterdam: North-Holland Publishing Company. Silver, Shirley and Wick R. Miller 1989 American Indian languages: Cultural and social contexts. Tucson, AZ: University of Arizona Press. Snapp, Allen, John Anderson and Joy Anderson 1982 Northern Paiute. In Studies in Uto-Aztecan grammar vol. 3: UtoAztecan grammatical sketches, Ronald W. Langacker (ed), 1-92. Arlington, TX: Summer Institute of Linguistics. Spaelti, Philip 1997 Dimensions of variation in multi-pattern reduplication. PhD dissertation, Department of Linguistics, University of California-Santa Cruz.

Reduplicative

allomorphy and language prehistory in Uto-Aztecan

349

Steriade, Donca 1988 Reduplication and syllable transfer in Sanskrit and elsewhere. Phonology 5: 73-155. Suärez, Jorge 1983 The Mesoamerican Indian languages. Cambridge: Cambridge University Press. Thomason, Sarah Grey and Terrence Kaufman 1988 Language contact, creolization, and genetic linguistics. Berkeley: University of California Press. Tuggy, David 1979 Tetelcingo Nahuatl. Studies in Uto-Aztecan grammar vol. 2: Modern Aztec grammatical sketches, Ronald W. Langacker (ed.), 1-140. Arlington, TX: Summer Institute of Linguistics. Urbanczyk, Suzanne this vol. Enhancing contrasts in reduplication. Voegelin, C. F. 1935 Tiibatulabal grammar. Berkeley: University of California Publications in American Archaeology and Ethnology vol. 34, No. 2. Whistler, Ronald 1992 Phonology of Sawai. Phonological studies in four languages of Maluku, Donald A. Burquest and Wyn D. Laidig (eds.), 1-32. Arlington, TX: Summer Institute of Linguistics. Whorf, Benjamin Lee 1935 The historical linguistics of Uto-Aztecan. American Anthropologist 37: 600-608. Whorf, Benjamin Lee and George L. Trager 1937 The relationship of Uto-Aztecan and Tanoan. American Anthropologist 39: 609-624. Wichmann, Soren 1995 The relationship among the Mixe-Zoquean languages of Mexico. Salt Lake City: University of Utah Press. 1999 On the relationship between Mixe-Zoquean and Uto-Aztecan. Kansas Working Papers in Linguistics 24.2: 101-113. Willett, Elizabeth 1982 Reduplication and accent in Southeastern Tepehuan. International Journal of American Linguistics 48.2: 168-184.

Reduplication in Tupi-Guarani languages: Going into opposite directions Frangoise Rose1

The aim of this paper is to investigate verbal stem reduplication in several Tupi-Guarani languages. The study reports mostly on Emerillon, for which the author collected field data in French Guiana, but also on other TupiGuarani languages: Wayampi of French Guiana (Grenand 1980, 1989; Ganozzi, p.c.)) Tupinamba (Rodrigues A. 1953), Kamaiura (Seki 2000), Brazilian Wayampi (Jensen 1989), Anambe (Juliäo, p.c.), Urubu-Kaapor (Kakumasu 1984), Chiriguano (Shuchard 1979) and Guarayo (Hoeller 1932) 2 . Tupi-Guarani languages in general display both a monosyllabic and a disyllabic reduplication pattern. However, while in other Tupi-Guarani languages the patterns copy the final syllables of the stem, in Emerillon the reduplication clearly affects the initial syllables. Let us note that up to now, general descriptions of Tupi-Guarani languages rarely included Emerillon data due to insufficient description. The reduplication process to be studied is a productive morphological process with the verb stem. Reduplication is also used with other parts of speech (like minor predicative categories, nouns or adverbs), but is really productive and frequent with verbs and ideophones only. Verbal reduplication is an expressive means for different nuances of aspect. Aspectual distinctions are otherwise expressed by verbal suffixes and particles in Tupi-Guarani languages. The first section will present the semantics of the two patterns, the second the phonological description of the reduplication patterns, both for Tupi-Guarani languages in general and for Emerillon in particular. The last one focuses on diachronic information that may explain the specificities of Emerillon as far as reduplication is concerned.

352 1.

Franqoise Rose Semantic analysis

The semantics of Tupi-Guarani reduplication is totally iconic: it conveys a meaning of multiplicity or repetition.

1.1. Dichotomy between monosyllabic and disyllabic reduplication Most descriptions of Tupi-Guarani languages present a very neat dichotomy between the semantics of monosyllabic and disyllabic reduplications, i.e. reduplications in which the copy consists of one and two syllables respectively, irrespective of the length of the reduplicated stem (Jensen 1990 and 1998 on Tupi-Guarani, Seki 2000 on Kamaiura, Rodrigues A. 1953 on Tupinamba and Jensen 1989 on Wayampi). Monosyllabic reduplication is generally said to express successive actions and disyllabic reduplication to have a frequentative or iterative meaning. Spontaneous examples of Emerillon make this semantic distinction stand out. The verbs kusug "wash" and hem "go out" both occur with each type of reduplication, with two different meanings. (1)

a.

o-ze-kusug3 3-REFL-wash 'She washes (herself).'

b.

"e-[ku]kusu(g)-katu arj 2SG.IMP-RED-wash-well D E M ' " W a s h these things well.'"

c.

o-itun-itun, o-[kusu] kusu(g)-katu-e?e, 3 -RED-smell 3 -RED-wash-well-INTENS pug o-ißu-ißurj lay.down 3-RED-put ' H e smells them, washes them well and puts them down each time.'

Emerillon

ba?e-kom" thing-PL

When only one syllable of the verb kusug is reduplicated (lb), the hero of the story is asked to wash a lot of dishes: it expresses successive action on all objects. In the example ( l c ) with disyllabic reduplication, the same hero does it frequently, every time his master goes away: it expresses iterativity.

Reduplication in Tupi-Guarani languages

(2)

a.

eiba o-hem bokal-a-wi o-/zoEmerillon pet 3-go.out jar-a-ABL 3-go 'The frog goes out of the jar.'

b.

amö kito-kom δ-[he] hem other frog-PL 3-RED-go.out 'Other frogs go out.'

c.

[öhe] ö-hem-ne o-?a RED-3-go.out-CONTRAST 3-fall 'He goes out again and falls down.'

353

In (2), whereas the example with monosyllabic reduplication (2b) shows that the different subjects realize the action successively, one frog going out after the other, the disyllabic reduplication (2c) presents a very repetitive action of a single actor, the hero, going in and out the monster's belly. Additional examples from different languages of the family show that the current definition of these functions in terms of succesivity VS iterativity is too restrictive, since they may vary according to the types of situation denoted by the verbs and according to the types of participants (especially with plural or mass nouns). We will therefore use Cusic's terms "eventinternal" and "event-external plurality" (Cusic 1981). This vocabulary, less simplistic, distinguishes "repetitive" action from "repeated" action. "Repetitive" action must be conceptualized as an "event-internal" verbal plurality, a series of iterated instances: "the units of action are conceived of as confined to a single occasion, and to a single event on that occasion." On the other hand, "repeated" actions must be conceptualized as an "eventexternal" verbal plurality: "the units of action are potentially distributable, though not necessarily distributed, over multiple occasions". This iterative meaning includes habitual and ongoing events.

1.2. Monosyllabic reduplication expresses event-internal repetition Monosyllabic reduplication is presented by Jensen (1989, p. 119) for Ancient Guarani, Kamaiura, Parintintin, Urubu and Wayampi as expressing successive actions: with intransitive verbs, the action is realized by a succession of subjects as in (3), and with transitive verbs, it is realized on a succession of objects as in (4). Rodrigues designs it as a "multiple realization of the process, either successive, either simultaneous".

354

Frangoise Rose

(3)

a.

oro-por

1 EX-jump 'we jumped'

(4)

b.

oro-[po]p0r 'we jumped, one after the other'

a.

a-i-mokon 1 SG-3-swallow Ί swallowed it.'

b.

a-i-mofko Jkon Ί swallowed one after the other.'

Tupinamba, Jensen 1990 my segmentation

Tupinamba, Jensen 1990 my segmentation

Previous examples of Emerillon showed successive repetition with all subjects (2b), and with all objects (lb). But with other verbs, monosyllabic reduplication does rather express internal repetition of the process, without implying the plurality of any participant. In the following example, the process of "gnawing" is necessarily made of several "bitings". (5)

äducfca wila o-[su]su?u rat wood 3-RED-bite 'The rat gnawed the wood.'

Emerillon

Closely related is the meaning of distribution of the result: in (5) the "gnawed" wood is in several pieces, just as in (6) a canoe is "split at different points". (6)

ial o-[ze]zeka canoe 3-RED-break 'The canoe split at different points'.

Emerillon

These examples illustrate exactly what Garcia-Medall (2003, p. 33) names "disintegration of the object"4 in his typology of reduplication in Native American languages. An analysis ä la Vendler (1967) shows that verbs usually denoting "achievement", as in (5) and (6), denote "accomplishment" once the reduplication has turned the basic situation into a non punctual process. This matches Grenand's analysis of Wayampi monosyllabic reduplicated forms as expressing "an action that lasts, while repeating"5 (Grenand 1980, p. 49). These last examples outmatch simple successivity, but they all stand in the semantic area of "event-internal" plurality.

Reduplication in Tupi-Guarani languages

355

1.3. Disyllabic reduplication expresses event-external repetition As for disyllabic reduplication, Jensen (1989, p. 120) presents it with a frequentative or iterative meaning. The action may be repeated several times, or in several places. (7)

(8)

a.

oro-por 1 EX-jump 'we jumped'

b.

o[ ropo ]ro-por ' w e j u m p e d frequently'

a.

a-i-mokon lSG-3-swallow Ί swallowed it.'

b.

a-i-[moko Jmokon Ί swallowed them frequently.'

Tupinamba, Jensen 1990 my segmentation

Tupinamba, Jensen 1990 my segmentation

Examples ( l c ) and (2c) of Emerillon also showed this meaning of iterativity. In those sentences, the verb (normally denoting an accomplishment or an achievement) expresses, in its reduplicated form, an activity that lasts a certain time without any modification. Chiriguano examples given by Schuchard (1979) suggest that the disyllabic reduplication pattern may convey a progressive meaning. Hoeller (1932) describes reduplication in Guarayo as having the following values: "often, continuously, little by little, more and more, at intervals, here and there" 6 . These different meanings can be subsumed under the general function of "event-external repetition". Up to now, we have argued that the dichotomy between monosyllabic and disyllabic reduplication patterns must be described in a less simplistic way than successive vs iterative meaning and should also take into account the type of situations denoted by the verbs and the types of participants. This may be done by using the larger opposition "event-internal" VS "eventexternal" plurality. Nevertheless, the dichotomy commonly presented for Tupi-Guarani languages is not so clearly cut in many languages of the family, including Emerillon.

356

Frangoise Rose

1.4. Some cases of disyllabic reduplication expressing event-internal plurality A good number of examples with disyllabic reduplication exemplify the "event-internal" repetition meaning, which is therefore not restricted to monosyllabic reduplication. Just as with monosyllabic reduplication, the meaning of internalrepetition can be due to the plurality of participants (subjects or objects) as in (9) or to the notion of a distributed result as in (10). In both cases, it seems that the denoted situation is an activity. In (9), the reduplicated form of cfyika "kill" denotes the traditional activity of killing fish that have been poisoned with the toxic juice of a creeper. (9)

mun-a-kom o-[cfyika Jckika-n. people-a-PL 3-RED-kill-PL 'People were killing them.'

Emerillon

In (10), the woman does not use her axe only once, but a succession of blows is necessary to form the activity of cutting wood for the fire. (10) wmwi-enam wila-we woman-TOP.SW wood-also 'The woman is cutting wood'.

o-[ eta]eta. 3-RED-cut

Emerillon

Descriptions of other Tupi-Guarani languages suggest that they also display bisyllabic reduplication with an internal-repetition meaning. Some Chiriguano examples given by Schuchard (1979) suggest an intensive meaning: (11)

e-i-fnupäj-nupä. 1 SG.IMP-3-RED-hit '"Hit him hard!'"

Chiriguano my segmentation

Grenand (1980) proposes another meaning for Wayampi: a weakened meaning. (12) a. b.

a-su?u Ί bite' a-[su?u]su?u Ί chew'

Wayampi, Grenand 1980 my segmentation

Reduplication in Tupi-Guarani languages

357

1.5. Summary In this semantic section, two facts have been underlined: -

First, the dichotomy between monosyllabic and disyllabic reduplication must be described in a less simplistic way than successive versus iterative meaning, and should also take into account the type of situations denoted by the different forms of the verb and the different types of participants. This can be done by using the larger opposition "event-internal" versus "event-external" plurality.

-

Second, in reality, the elegant semantic dichotomy is not easily applicable to the Emerillon data, since "event-internal" plurality may also be expressed by disyllabic reduplication. This is true for many other Tupi-Guarani languages and will be explained in the diachronic section of this paper.

But before this, we must examine the main difference in reduplication between Emerillon and other Tupi-Guarani languages, which concerns the phonological patterns involved in reduplication.

2.

Phonological description

2.1. Reduplication patterns We now turn to a description of reduplication as a phonological operation. Examples from all Tupi-Guarani languages suggest that the reduplicant copies the base according to a mono or disyllabic template. Only one copy is made, and no other modification alters the stem. (13)

a.

a-i-mokon 1 SG-3-swallow Ί swallowed it.'

Tupinamba, Jensen 1990 my segmentation

b.

a-i-mofko ]kon Ί swallowed them one after the other.'

c.

a-i-[ mokojmokon Ί swallowed them frequently.'

358

Frangoise Rose

Canonical forms of Tupi-Guarani morphemes consist of one or two syllables (rarely more), each vowel being a syllable nucleus. Closed syllables (with a coda consonant) are restricted to the final position in the morpheme. This final coda consonant is excluded by the reduplication process. (14)

a.

-pyhyk 'take, hold'

b.

-pyhy-pyhyk 'take several times'

Kamaiura, Seki 2000 my segmentation

A morphophonological explanation could be put forward: at a morpheme boundary, one consonant preceding another tends to be deleted. In other words, if two identical strings follow each other, the final consonant of the first string must be dropped in front of the initial consonant of the second. But the following example shows that this hypothesis is irrelevant: the elision of the consonant appears also in front of a vowel. (15)

a.

o-?al 3-fall ' H e falls'

b.

*

c.

[o?a ]o?al

Emerillon

[o?al]o-?al

We suggest that the absence of final consonant in the reduplicant may be explained in terms of moras: the reduplication process in Tupi-Guarani languages creates reduplicants consisting only of monomoraic syllables. The final consonant is thus systematically excluded from the copy. Reduplication operates on a base consisting of the verbal stem or part of it, excluding suffixes (ΤΑΜ, number) or clitics ("particles"). (16)

[o?a]o-?al-oij RED-3-fall-PL 'They fall one after the other.'

Emerillon

A logical problem is raised by disyllabic reduplication on monosyllabical stems. Tupi-Guarani languages, including Emerillon, all seem to solve the problem the same way: to fill the disyllabic reduplicant template, they

Reduplication in Tupi-Guarani languages

359

extend the base by including the preceding syllable, which is the last or single syllable of either a person marker, or a voice prefix. (Verbs are usually preceded by a person prefix in these languages.) (17)

(18)

a.

ere-syk 2SG-arrive 'You arrive.'

Tupinamba, Rodrigues 1953 my segmentation

b.

e[resv]re-syk 'You frequently arrive.

a.

a-lo-wag 1 SG-CAUS.COM-go Ί moved the pot.'

b.

a-flowajlo-wag pol Ί moved the pot several times.'

pol pot

Emerillon

Thus the phonological string of the copy does not correspond to a unique morphological unit. In summary, reduplication in Tupi-Guarani languages involves different domains, defined on morphological and phonological criteria: -

either one syllable of the verbal stem (all of it, if it is monosyllabic), any possible coda consonant being excluded two syllables of the verbal stem (all of it, if it is disyllabic), any possible coda consonant being excluded one syllable of the verbal stem plus the last syllable of the preceding m o r p h e m e (with monosyllabic stems only), once again any possible coda consonant being excluded

The interesting fact is that these domains do not include the same syllables of the stem in Emerillon and in other Tupi-Guarani languages. Let us now illustrate the t w o patterns (monosyllabic and disyllabic) for both TupiGuarani in general and Emerillon in particular.

360

Frangoise Rose

2.2.

Monosyllabic reduplication

The first reduplication template is monosyllabic, one could also say in that case monomoraic, the last consonant being deleted.

2.2.1. In most Tupi-Guarani languages For the languages I know of, the monosyllabic pattern reduplicates the final syllable in Wayampi, Tupinamba, Guarayo, Anambe and Kamaiura. (19)

(20)

a.

o-sala 'he breaks'

b.

o-sa[la]la 'he splits'

a.

a-piso. Ί pull out a feather'

b.

a-pifso]so Ί am plucking a fowl'

Wayampi, Grenand 1989 my segmentation

Wayampi, Grenand 1980 my segmentation

2.2.2. In Emerillon On the contrary, monosyllabic reduplication in Emerillon copies the first syllable of the stem. Several examples with plurisyllabic (all disyllabic) stems show this clearly: (21)

"fre, wajie arj o-[pi]pinö" brother incessantly DEM 3-RED-fart '"Brother, this one keeps on farting'"

Emerillon

Reduplication in Tupi-Gnarani languages 2.3.

361

Disyllabic reduplication

The second reduplication pattern is disyllabic, with monomoraic syllables, the coda consonant being excluded. 7

2.3.1. In most Tupi-Guarani

languages

In parallel to the monosyllabic pattern, the disyllabic pattern reduplicates the last two syllables in Wayampi, Tupinamba, Anambe, Chiriguano, Guarayo and Kamaiura. (22)

(23)

a.

o-j-apisi 3-REFL-fight 'They fight.'

b.

o-j-a[pisijpisi 3-REFL-RED-fight 'They (kids) quarrel.'

a.

a[ikä] 'to trim'

b.

a[ ikä]ika 'to trim several times'

Wayampi, Ganozzi, p.c. my segmentation

Anambe, Juliäo, p.c. my segmentation

2.3.2. In Emerillon The Emerillon disyllabic pattern involves two syllables of the verbal stems. All examples being with monosyllabic or disyllabic stems, it is actually impossible to decide if they are the initial or final syllables of the stems. (24)

a-[ ißu Jipuy zebi?a 1 SG-RED-put music Ί am always listening to music.'

Emerillon

362

Fr anweise Rose

2.4. Discussion The oddity of the different anchorage of reduplication within the TupiGuarani family seems difficult to explain. Moreover, Emerillon functions differently from the languages of the same subgroup (8th subgroup, in Rodrigues 1984-85) presented in this paper: Wayampi and Anambe8. Reduplication patterns involve the same kind of canonical forms in all Tupi-Guarani languages, but they copy different parts of the stem. Nevertheless, with disyllabic reduplication on disyllabic stems, the forms resulting from the two different models of reduplication do not differ (25). (25)

a.

o-nupä-nupä 3-hit-RED 'He clubbed it (a fish) repeatedly.'

b.

o-nupä-nupä 3-RED-hit 'He hit it several times.'

Wayampi, Jensen 1990 Jensen's segmentation Emerillon

A further discussion to be considered at this point is how to tell apart the reduplicant from the base. We assumed in this paper that the reduplicant was placed before the base. An alternative analysis is more widespread: the reduplicant would copy one or two full syllables (including the coda consonant) and follow the base, which would itself lose its coda consonant (Rodrigues 1953 for Tupinamba, Jensen 1993 for Wayampi & 1998 for Tupi-Guarani in general, Everett and Seki 1985). 0 is an example the way Jensen presents it: (26)

a. a-i-mokon 1 SG-3-swallow Ί swallowed it.'

Tupinamba, Jensen 1990

b. *o-i-moko-kon Proto-Tupi-Guarani, Jensen 1998 3A-3P-swallow-REDUP Jensen's segmentation 'He swallowed them one after the other.' In this paper, I argue that this description is needlessly complicated: why would the coda consonant be deleted from the base, while it is saved in the copy? The description gives no explanation for this (remember we saw with (15) that it could not be a morphophonological rule). Everett and Seki (1985) still argue for a morphophonological rule. They consider, for

Reduplication

in Tupi-Guarani languages

363

Kamaiura only, than the reduplicative morpheme is expressed as a /CVCVC/ suffix. When the phonemic melody is too short to fill the initial C slot of the suffix template, the consonant deletion rule is still triggered by the empty consonant. We have then to accept that the deletion rule operates at the prosodic template level, rather than on a phonemic melody. My preference goes for a more economical analysis, using a reduplicant template based on prosody: if each reduplicated syllable must be monomoraic, then the coda consonant is logically absent from the copy, but still present in the base. Consequently, we assume that the copy precedes the base. We understand that the kind of description illustrated in (26) -suffixation rather than prefixation- was motivated by the wish to maintain correspondence between the phonological and morphological boundaries. Here is one of Rodrigues' examples, with his own segmentation: (27)

a.

ere-syk 2SG-arrive 'You arrive.'

b.

ere-sy-re-syk 'You frequently arrive.'

Tupinamba, Rodrigues 1953 Rodrigues' segmentation

The segmentation tries to preserve entire morphemes. Nevertheless, the reand sy- parts do not correspond to a whole morpheme. In their article, Everett and Seki (1985) reject Bruce Hayes's account of Kamaiura reduplication as the infixation of a -CVCV- skeleton before the penultimate syllable of the stem, underlining the fact that there is no infixation9 in TupiGuarani. As a counter-argument, note that if we adopt their analysis of suffixation for the Emerillon data, once again reduplication will look like "infixing", or rather like internal reduplication. The copy, if it follows the base, shows up within the stem (28b). (28) prefixation (our analysis of Emerillon): a. o-f su ]su?u (29)

suffixation (following Everett & Seki analysis of Kamaiura): b. o-su[su]?u

Since the data make it necessary to dissociate phonological and morphological segmentation, we will prefer the simpler hypothesis concerning

364

Frangoise Rose

Copy-Base order: the copy precedes the base, and the absence of the coda consonant in the copy is explained by its canonical form consisting of only monomoraic syllables. Thus the segmentation of (26) and (27) is for us:

(30) a.

b.

(31) a.

b.

a-i-mokon 1 SG-3-swallow Ί swallowed it.'

Tupinamba, Jensen 1990 my segmentation

a-i-m ofko Jkon Ί swallowed them one after the other.' ere-svk 2SG-arrive 'You arrive.'

Tupinamba my segmentation

e [resy]re-syk 'You frequently arrive.'

The copy is placed just before the base, therefore inserted within the lexeme in (29) and within the person marker in (30). This case just reminds us that morphological processes are not limited to juxtaposition of morphemes. In Optimality Theory, it could be nicely explained in terms of competition between constraints. The markedness constraint - requiring only monomoraic syllables in the copy - is higher in the hierarchy than the faithfulness constraint - requiring correspondence between output and lexical input (see for example Kager 1999). The different analysis of the order Base/Copy does not nevertheless solve the problem of the unusual anchorage of Emerillon reduplication within the Tupi-Guarani family.

3.

Diachrony

In the preceding sections, it has been established that Emerillon does not follow the semantic and phonological models of reduplication of other Tupi-Guarani languages. This difference may partly be explained by diachronic or areal phenomena. Note first that descriptions do not explicitly claim to give "reconstructions" for Tupi-Guarani languages, but rather use Rodrigues' description of Tupinamba reduplication as a basis and extend it to other languages.

Reduplication in Tupi-Guarani languages

365

3.1. Diachronic information explaining the semantics of Emerillon reduplication patterns Diachronic information helps us understand the blurred semantics of Emerillon reduplications. The most interesting evolution, often left out, is that monosyllabic reduplication as a productive phenomenon has in fact disappeared from many Tupi-Guarani languages. These languages retain frozen forms, like mokö-kö ('swallow one after the other') in Wayampi, but do not have monosyllabic reduplication as a productive phenomenon. In such languages, disyllabic reduplication takes on the two meanings, e.g. in Wayampi. In Emerillon, occurrences of monosyllabic reduplication are clearly less numerous than those of disyllabic reduplication. The semantics of monosyllabic reduplication is more homogenous than that of disyllabic reduplication. One hypothesis is that monosyllabic reduplication in Emerillon is beginning to decline, and disyllabic reduplication takes on the meaning of "event-internal plurality" as a possibility.

3.2. Diachronic or areal information about the Emerillon phonological patterns of reduplication Concerning the phonological pattern, no clear explanation emerges from diachrony. Three points are however worth mentioning: Interesting is Jensen's assertion that monosyllabic reduplication repeats the last stressed syllable of the word, and disyllabic reduplication the last stressed syllable and the one preceding it (Jensen 1989, p. 119-21). However, if stress in Proto-Tupi-Guarani is reconstructed on the final syllable of the stem, it occurs on the penultimate syllable in several languages like Chiriguano and Wayampi. In those languages, reduplication still involves the last syllable or the last two syllables. Position of stress does not seem to constitute a good explanation for the different models of reduplication between Emerillon (where stress seems to occur either on the final or on the penultimate syllable of the stem) and the other Tupi-Guarani languages. Second, in Xipaya, a Tupi language of the Juruna family, initial reduplication is well attested (Rodrigues C., p.c.). Xipaya's reduplication involves both ends of the verb stem. In the "real" mode, the initial syllables are copied, while in the "unreal" mode, the last syllables of the verb are copied. Therefore, Emerillon is not the only Tupi language with initial reduplication; unfortunately, we do not know the extent of this phenomenon in the Tupi family.

366

Frangoise

Rose

Finally, in A r a w a k , another large A m a z o n i a n family, initial and final reduplication coexist, sometimes in the same languages. (Aikhenvald 1996) If this does not tell us why Emerillon displays an initial reduplication, at least it s h o w s that Emerillon is not such an oddity in the region.

Abbreviations 1EX 2SG.IMP ABL CAUS.COM CONTRAST DEM INTENS PL RED REFL TOP.SW

first person exclusive second person of the imperative ablative postposition causative-comitative contrastive demonstrative intensive plural reduplicant reflexive topic switch

Notes 1. 2. 3.

4. 5. 6. 7. 8.

9.

I am grateful to Denis Creissels for his comments and support. Concerning these last two languages, I wish to thank Wolf Dietrich for sharing the information. Thanks also to Risoleta Juliäo for Anambe's data. Transcriptions follow the IPA conventions, except those of Tupinambä and Kamaiurä where " / ' stands for "i". The accent on a vowel indicates stress. Underlined strings correspond to the base of reduplication, and brackets mark the reduplicant. The analysis of strings as base and reduplicant is ours, except in part 2.4 where this point is discussed. My translation for "desintegracion objetual". My translation for "une action qui se prolonge, qui dure, tout en se repetant". My translation. The only four possible patterns are V.V, V.CV, CV.V, CV.CV. Information on Urubu (also from the 8th subgroup) is limited to examples in Kakumasu's work. His examples are all monosyllabic reduplication on monosyllabic stems, or disyllabic reduplication on disyllabic stems, which does not inform us on the initial/final reduplication discussion. Infixation is to be understood in terms of affixation of a bound morpheme within a stem.

Reduplication in Tupi-Guarani languages

367

References Aikhenvald, A. Y. Amazonian languages. The Third Australian Linguistic Institute, 1996 Australian National University. Cusic D. 1981 Verbal plurality and aspect. PhD diss., Stanford University, cited in "Reduplication and infixation in Yurok: morphology, semantics and diachrony", Garrett Α., 2001, IJAL, Vol. 67, n° 3. Everett, D. L. and Seki L. 1985 Reduplication and CV Skeleta in Kamaiurä. Linguistic Inquiry 16 (2): 326-330. Garcia-Medall, J. Sobre reduplicacion morfologica en lenguas amerindias. In 2000 I Simposio Antonio Tovar sobre Lenguas Amerindias, E. Ridruejo and M. Fuertes (eds.). Valladolid: Tordesillas. Grenand, F. 1980 La langue wayäpi (Guyane franqaise). Phonologie et grammaire. (Langues et civilisations ä tradition orale 41.) Paris: SELAF. Dictionnaire wayäpi-frangais. Paris: Peeters / SELAF. 1989 Hoeller, A. 1932 Grammatik der Guar ay o-Spr ache. Hall (Tirol). Jensen, C. 1989 Ο desenvolvimento historico da lingua Wayampi. Campinas: Editora da Unicamp. 1990 Cross-referencing changes in some Tupi-Guarani languages. In Amazonian Linguistics, Studies in Lowland South American Languages, D. Payne (ed.), 117-158. Austin: University of Texas Press. Comparative Tupi-Guarani Morpho-syntax. In Handbook of Amazo1998 nian languages, D. C. Derbyshire and G. K. Pullum (eds.), Vol. IV, 490-603. Berlin/New York: Mouton de Gruyter. Kager, R. 1999 Correspondance in reduplication. In Optimality Theory, R. Kager. Cambridge: Cambridge University Press. Kakumasu, J. 1984 Urubu-Kaapor. In Handbook of Amazonian languages, D. C. Derbyshire and G. K. Pullum, Vol. I, 326^103. Berlin/New York: Mouton de Gruyter. Rodrigues, A. D. 1953 Morfologia do verbo Tupi. Letras, Separata n°l, 121-152, Curitiba. 1984-85 R e d o e s internas na familia lingui'stica Tupi-Guarani. Revista de Antropologia 27-28: 33-53.

368

Franqoise

Rose, F. 2000

Rose

Elements de phonetique, phonologie et morphophonologie l'emerillon (Teko). Master Thesis, Universite Lyon 2.

de

Schuchard, B. 1979 Nane fie. Gramdtica guarani para castellanohablantes. Santa Cruz de la Sierra. Seki, L. 2000 Gramdtica do Kamaiurd. Campinas: Editora da Unicamp. Vendler, Z. 1967 Linguistics and Philosophy. Ithaca: Cornell University Press.

On the borderline of reduplication: Gemination and other consonant doubling in Arabic morphology1 Dina El Zarka

1.

Introduction

Gemination and reduplication in Arabic morphology have been classified both as essentially the same thing and as two different procedures. While morphological gemination is a comparably uncontroversial term in linguistic literature, reduplication, its definition and delimitation from other repetition and doubling phenomena is still a matter of some debate. One such delimitation problem concerns the differentiation between reduplication and recursive structures in syntax, another problem of delimitation arises at the other end of the continuum, between morphology and phonology, involving the doubling of bare segments or the "tiny bits", as dubbed by Inkelas (cf. Inkelas, this volume). This paper is concerned with the latter, drawing on data from Arabic verbal morphology. It is assumed that in Arabic, and in Semitic in general, bare consonants can be doubled to achieve some change in meaning. The Arabic verbs investigated here exhibit the doubling of one or two consonants, like farfah (C1VC2C1VC3) or bargog (CWC2C3VC3.) and laflif (C1VC2C1VC2), or the most frequent type of doubling, the doubling of C 2 , which results in gemination, like in kassar or fakkar (C1VC2C2VC3). Trying to decide whether this phenomenon should be viewed as reduplication or not will inevitably lead us to some basic issues that have been dealt with in the recent linguistic debate. (i) (ii)

Is reduplication of bare segments only a phonological process or does it bring about semantic change? Does reduplication necessarily obey prosodic requirements following the assumptions of Prosodic Morphology (McCarthy and Prince 1986; Marantz 1982) or can the doubling of bare segments be counted as an instance of reduplication?

370

Dina El Zarka

(iii)

Is gemination and other segmental doubling in Arabic to be considered as reduplication or spreading (McCarthy 1981, 1982)? Is the strong grammaticalization hypothesis that predicts a grammaticalization path along the following line: full reduplication > partial reduplication > [...] > gemination (Niepokuj 1997: 63) correct?

(iv)

The functional question (i) does not seem to be very problematic: reduplication as a grammatical phenomenon is commonly held to involve a change in meaning, thereby excluding phenomena such as the well-known reduplicative constructions that occur in babbling, consonant harmony and long consonants as in Italian 'raddoppiamento sintattico'. The formal problems (ii, iii), however, seem to be far from being settled. As far as question (iv) is concerned, the Arabic data will provide further evidence against the strong grammaticalization hypothesis.

2.

Preliminaries

2.1. The mechanism of Semitic morphology In Arab grammatical tradition Arabic morphology has always been described as a system of roots (djiör) and patterns (wazn) which are filled out by the intercalation of vowels to the consonantal root skeleton. This model was established systematically in modern linguistics with John McCarthy's dissertation (1979), which was the starting point of a prosodic theory of non-concatenative morphology (McCarthy 1981). This view has been challenged in recent linguistic work, which attempts an analysis without roots and patterns, or only without roots, e.g., Bat-El (1994) for Modern Hebrew and Ratcliffe (1998 and prior work) for Arabic. This is not the place to discuss this issue. The necessity of assuming word-based derivation in many cases of Arabic word formation, like in the formation of the "broken" plural (Ratcliffe 1998) does not, in my opinion, obliterate the traditional analysis of Arabic morphology as a root-pattern-system but rather enriches it.2 I believe the root exists as an, admittedly abstract, lexical morpheme that provides the basic consonantal skeleton for derivational processes which are considered to be an interplay of the root, vocalic melodies and certain prosodic patterns. Sometimes they also involve other morphological processes like doubling or affixation. The template is viewed as a kind of constraint on the other morphological operations (cf. also Ratcliffe 1998: 4 2 50).

On the borderline of reduplication

371

Arabic roots are mostly triliteral (henceforth cited between and the most common simplex verb contains only three consonants with interspersed vowels, and exhibits a pattern CVCVC or katab (following the Arab tradition of using an existing root to illustrate the pattern). This pattern conveys the basic verbal meaning of the root. All other verbal classes are derived by augmentation, either with a morphological affix or by the doubling of consonants. This system in principle holds true for Classical Arabic (CA) as well as Modern Standard Arabic (MSA) and for the NeoArabic varieties spoken today. The derived forms add something to the basic meaning of the root, or alter it in some way, and thus constitute semantic verbal classes.

2.2. Gemination and reduplication in former studies In 1862 August Friedrich Pott formally divided doubling into two sub-procedures, gemination and reduplication. Gemination, contrary to the meaning of the modern term, is viewed as total doubling ("Wiederholung im Ganzen") and reduplication as partial doubling ("verkürzte und nur zum Theil, also bloß andeutungsweise vollzogene Wiederholung", p. 16). Reduplication is further divided into a) strengthening of bare segments ("bloße Steigerung von Einzel-Lauten", p. 16) and b) reduplication proper ("eigentliche] Reduplikation", p. 18). This last distinction is obviously based on the shape of the reduplicant. In case bare segments are strengthened, the reduplicant does not project a syllable of its own. Pott explicitly cites vocalic lengthening as "vowel strengthening" and the gemination of the second radical in Semitic languages as "consonant strengthening" (p. 17). If it is only one segment that is doubled, in cases of reduplication proper, the result is always a syllable in its own right. When initial consonants are strengthened, they are said to have an accompanying vowel as in Sanskrit perfect formation (p. 18). Considering that Pott's survey subsumes all different kinds of repetition phenomena under the heading "Doppelung", whether they are grammatical or extra-grammatical, it comes as no surprise that he should include every kind of formal doubling as well. But also Edith Moravcsik (1978: 309), in her seminal article on reduplication, explicitly counts Syrian Arabic gemination of C2 as an instance of reduplication. Consonantal and vocalic doubling is also viewed as a kind of reduplication in more recent surveys on reduplication (Rubino, this volume). The opposite view has been expressed, for instance, by Wiltshire and Marantz (2000). They exclude gemination in Semitic from reduplication (p. 558) and group it together with consonant

372

Dina El Zarka

mutation phenomena (p. 567), but do not bring forth any arguments to sustain their claim. While specialists on Arabic or Semitic languages mostly put all doubling phenomena of Arabic together under the heading of reduplication (or doubling) like Ibrahim (1982), the history of generative treatments is one of keeping things apart (in the framework of autosegmental and metrical phonology), but recently also of aiming at a unified treatment within Correspondence Theory (Imouzaz 2002). The arguments put forward are mostly purely formal and theory-internal and to the greater part neglect the semantic dimension. In these accounts, gemination has been viewed variously as multiple association of root segments in the framework of autosegmental phonology (McCarthy 1979, 1981) and within the theory of metrical phonology (Angoujard 1988), or as the addition of a mora within Prosodic Morphology (McCarthy 1992), while other doubling has been attributed variously to (long distance) consonant spreading or reduplication (Broselow and McCarthy 1983), depending on the adjacency of the segments. More recently, with the advent of Optimality Theory, a different approach has been put forward and the distinction between long distance consonant spreading and reduplication has been challenged (Gafos 1995, 1998; Rose 2000). OT-analyses of different languages have come to treat all kinds of bare consonant repetition as being essentially the same, like Hendricks (1999) and Imouzaz (2002).

2.3. Reduplication as a morphological operation Reduplication in Semitic is not a process that adds material to the base in a purely concatenative manner; it rather obeys the requirements of template satisfaction. This is the main reason why such doubling phenomena have often been analyzed as the spreading of phonetic information. In line with the approach to Arabic morphology as outlined above, I argue in the following section that bare consonant doubling in Arabic is a morphological means of word formation. In morphological terms, reduplication is the repetition of all or part of a morphological entity (root, affix, stem). If the consonantal root is considered as a genuine morphological entity, there is no problem in classifying the repetition of part of that root as reduplication. If we have a look at the literature on reduplication, we however observe a certain reluctance to classify consonant doubling as such. I assume that the reason for this reluctance lies in what we expect prototypical reduplication to look like. Prototypical reduplication is supposed to be highly iconic

On the borderline of reduplication

373

(full reduplication) and morphotactically transparent. Full reduplication is easier to perceive, to memorize and evaluate, but even adjacent reduplicated syllables are easily recognized as being the same - much less so, if only singleton segments are doubled. So it seems to be a matter of the phonological make-up, the prosodic structure and of adjacency, whether we easily perceive certain structures as reduplicated. In phonological terms, reduplicated structures should preferably project syllables on their own with the segments showing the same sequence in the reduplicant as in the base. Such a preference is paralleled by the extralinguistic principles of poetry, Alliteration and Rhyme, as suggested by Yip (1999). Both, the linguistic and poetic principles can be further deduced from general natural principles of rhythmicity and gestalt preferences (cf. Dressier 1996). Semiotically, adjacency facilitates indexicality (Dressier 1996), hence alliteration only counts when the phonemes are at a sufficiently close distance (Dressier 1990). Thus, prototypical reduplication is supposed to be adjacent. This takes us back to Pott's classification of the reduplication phenomena, which divides reduplication into "proper" reduplication and the "strengthening" of segments. I therefore assume reduplication that abides by the principle of alliteration and rhyme (e.g., syllable reduplication) to be more natural than reduplication of bare segments. I will return to this point in section 4 after I have discussed the Arabic data.

3.

The Data

3.1. Form and meaning of Arabic CVCCVC-verbs The data presented here will exclusively be tokens of the second disyllabic pattern (CVCCVC), mostly from Neo-Arabic vernaculars, but also from Classical (CA) and Modern Standard Arabic (MSA) which is the modern continuation of the Classical variety, especially as far as morphology is concerned. 3 Arabic verbs of the CVCCVC-pattern can have the following abstract segmental shapes: katkat (with a reduplicated root or syllable) katkab (with a doubled Ci) kattab (with a doubled (i.e., geminated) C2) katbab (with a doubled C 3 ) C1C2C3C4

374

Dina El Zarka

With regard to semantics, the verb forms can be said to derive from an existing form, be it a simplex verb - as it is in most cases - or a noun. But not every augmented pattern can synchronically be traced back to a simplex form, augmented lexical forms do not seem to be dependent on the prior existence of simplex ones. As mentioned above, I assume that formally the derivational process utilizes the root, and pairs it with a certain pattern to arrive at the desired lexical form.

3.1.1. Patterns with gemination Stem II of the verb is derived by the doubling of the second consonant, which is commonly called 'gemination' in Arabic philology, and is thus the Arabic realization of a common Semitic D(oubling)-stem. Wright ([1896] 1974/1: 31) describes the original meaning of the Araita^-pattern in CA as intensive or "extensive", thus translating the old Arab grammarians' categorization of muba:laya 'exaggeration' and takOi:r 'augmentation, multiplication' which means that an action is being carried out for a longer period of time (continuous) or repeatedly (iterative, frequentative) or by a number of people or in a number of places (distributive). In a tentative account of the functions of the kattab-stem, Greenberg (1991) noted that the Semitic "Intensive" might better be analyzed as expressing verbal plurality (cf. Dressier 1968; Cusic 1981). Table (1) gives some examples of the different intensive/pluractional meanings of the kattab-form: (1) root

verbform

gloss

language

°frq° °frq °+ GEM

faraq farraq

to separate to disperse

CA/MSA CA/MSA

°drb° °drb°+ GEM

ctarab ctarrab

to beat to beat violently

CA/MSA CA/MSA

°ksr° °ksr° + GEM

kasar kassar

to break to break into pieces

CA/MSA CA/MSA

°bky° °bky°+ GEM

baka: bakka:

to weep to weep much

CA/MSA CA

°brk° °brk°

barak(a4 l-djamalu) barrak(a n-naVamu)

(the camel) kneeled down (the whole drove of camels) kneeled down

CA CA

On the borderline of reduplication

375

As the above table shows, the pluractional function of the kattab-pattern seems to have been partly lost over time. The dictionary of Hans Wehr (1977) on Modern Standard Arabic (MSA) gives for bakka: (°bky°+GEM) and barraka (°brk°+GEM) only the translations 'to make cry' and 'to make (the camel) kneel', respectively. This leads us to a second function of the &aita&-pattern, viz. the causative/factitive function, which is the dominating function in Neo-Arabic and MSA. A third, very common function, again as well in MSA as in Neo-Arabic varieties, is the denominative function which also extends to loan words. In Neo-Arabic and in MSA the causative and denominative functions can be said to be fairly productive, the pluractional/intensive meaning is still maintained by a considerable number of verbs. These functions have already developed in CA (Wright [1896] 1974/1: 31). But the extensive use of this pattern in Neo-Arabic has been furthered by the loss of stem IV (Paktab) of CA for factitive or causative signification, and its merger with stem I, with the second form taking over its function in many cases.

3.1.2. Patterns with full

reduplication

Traditional Arabic grammars like Wright ([1896] 1974) subsume the remaining CVCCVC-verbs under the heading quadriliteral, not differentiating the nature of the consonants. From a semantic point of view, this seems useful. Moreover, these verbs are also structurally related. Thus a quadriliteral verb may just as well consist of four as of three or two different consonants, with one or two of them being doubled. The most common of these is the katkat-pattern, a pattern with full reduplication of a biconsonantal (in some cases pseudo-) root, which results in the surface representation of two identical or nearly identical syllables. Words with that pattern can be found in all varieties and are often of onomatopoetic origin. Despite their extra-grammatical formation, these also fit into the overall derivational system, and the root can be extracted and made the basis for the derivation of nominal forms, like zalzal 'trembling (earth)' - zilza:l 'earthquake' (CA/MSA), was was 'whisper' - waswa:s 'whisperer (devil)' (CA/MSA), ta?ta? 'lisp' - ta?ta?a 'lisping' (Egyptian Arabic, henceforth EA). The origin of many of the biradical roots is not at all clear, and they do not seem to have common Semitic ancestors. This is why Moscati et al. (1980: 130) regard all quadriliteral forms in Semitic as innovations. Prochäzka (1995) gives a survey of the semantic properties of the reduplicated biradical roots and roughly divides them into onomatopoetic words

376

Dina El Zarka

that express acoustic phenomena like different voices and sounds, movements and other things that shall be neglected here. Even if many of these words are of onomatopoetic origin, words for different voices like snoring, rattling etc. and movements like shivering, shaking, swinging to and fro etc. share one common feature, namely the plurality of the action or iterativity of the event. 5 It has often been observed that Neo-Arabic varieties exhibit a strong tendency towards the use of quadriliteral verbs. While the doubling of C2 (gemination) is found all throughout Semitic (Kienast 2001: 227, Lipmski 2001: 390), the reduplication of biradicals is a feature of West Semitic (Brockelmann 1908: 520) and is especially common in Ethio-Semitic languages (Höfner 1951: 103, Unseth 2003: 268). Lipinski (2001: 414) mentions their frequent occurrence in Libyco-Berber. A preliminary investigation of Hinds and Badawi' dictionary of Egyptian Arabic (1986) clearly shows that, apart from obviously onomatopoetic words which do not show a corresponding simplex form, fully reduplicated verbs only arise if the root is composed of only two different consonants, thus I assume them to be derived from these roots, which are interpreted as bi-consonantal. There is no biconsonantal reduplication within triliteral verbs, as in that case relevant phonological material would have to be dropped to fit the CVCCVC template, which serves as a constraint to the morphological operation of reduplication. It seems that the younger katkat-vsrhs cover the same semantic range as the older kattab forms, i.e., from different kinds of pluractional and intensive significations to causativization (transitivization) and even intransitivization. Sometimes kattab- and katkat-forms are both attested, (mostly) with a differentiation in meaning, sometimes only one of the two forms can be found. There seems to be a tendency for the fully reduplicated forms to move towards stronger pluractionality in the sense of event-internal repetition, i.e., that one event at one occasion consists of repeated phases, like bite and nibble. Table (2) shows some examples from EA (examples from Moroccan Arabic (MA) are marked as such), which do not give an exhaustive overview of all different kinds of meanings these words, can take on:

On the borderline of reduplication

377

(2) root

simplex

geminated

reduplicated

°lfP

laff'wrap, wind, twist, coil'

laffif 'wrap, wind, twist, coil repeatedly'

Iaflif 'wrap up'

°lmm°

lamm 'collect, gather'

°Jmm°

Jamm 'smell'

Jammim 'make, let smell' JamJim 'sniff'

/gmm (MA) 'id.'

Jsmmam (MA) 'id.'

lamlim 'gather up, gather together'

JsmJsm (MA) 'id.'

W

mass 'suck'

masmas 'chew, suck on'

°bss°

ba&9 'look'

basbas 'ogle, make eyes, leer'

°dJJ°

daJJ 'mash, pound' dajjij 'mash or pound lightly'

dajdij 'reduce to fragments, smash, shatter'

°d??°

da?? 'pound, ham- da??a? 'go into detail, be da?da? 'crush, grind by mer, nail etc.' meticulous, scrutinize' pounding, etc.'

3.1.3. Patterns with doubled Ci or C3 and C/ FC\C?VC 4 According to Brockelmann (1908: 517), final doubling which results in the katbab-v&i\Qm in Arabic CVCCVC-forms is used in all Semitic languages to form iteratives. Three of the now mostly obsolete forms of Old Arabic (=CA) (stem IX, XI, XIV) are built according to that principle. Brockelmann does not mention the occurrence of CVCCVC-disyllables of that type in Old Arabic, but only gives examples from Maghrebinian (Neo-Arabic) varieties (Brockelmann 1908: 518; cf. also Kienast 2001: 236). A fair number of Neo-Arabic verbs show a doubled Ci after C2, which results in katkab. If doubling of C] were adjacent, the resulting pattern should be *kaktab which is not attested, obeying a general prohibition of Cj=C2 in Arabic roots (Greenberg 1950). It has variously been suggested that such forms are the result of dissimilation of the first part of a totally reduplicated form, such as in farfaba < tabiaba 'he gurgled' (Lipihski 2001: 414). This pattern is especially common in the Levantine varieties (cf. Cowell 1964: 11 Of.). One of the striking features of the above mentioned patterns, and likewise the pattern with four different Cs, is the fact that C 2 is significantly

378

DinaEJZarka

often represented by a sonorous consonant (r,l,n or a semivowel) which splits an original doubled C 2 or deconstructs a fully reduplicated form, with the sonorant mostly occupying the coda position of the first syllable: farqaT < faqqaΐ 'explode' (Brockelmann 1908: 510) or the already mentioned iariab < iabiab 'he gurgled'. Examples of such forms are legion, not only in the CVCCVC-pattern; the now obsolete XIIth stem was also probably built according to the same principle: ?ixdawdar Jihmarra or even the formation of stem I from a root °mrr° *marara > marra8. One possible argument against viewing the geminate as an instance of doubling is the fact that it cannot be split in the inflectional paradigm, as in the following CA-verb forms: (5) (i) (ii) (iii)

root °mrr° (stem I) °hmr° (stem IX,+RED(C 3 ) °ktb° (stem II, RED (C 2 )

3.sg.m.: +a marra ?ihmarra kattaba

l.sg.: +tu marartu ?ihmarartu kattabtu

The reason why the geminate is not split in (iii), is its position between two vowels, whereas in examples (i) and (ii), forming a geminate would yield an illicit consonant cluster as it is followed by a consonant: *marrtu, *Tihmarrtu? Geminated words structurally behave like all other CVCCVCforms. The decisive argument for viewing geminates as doubled, and not lengthened consonants, is the fact that their first segment can be replaced by another segment (cf. below). In the fully reduplicated forms the root consists of only two different consonants, both of which are doubled. This amounts to repeating the whole syllable /CiVC 2 /. 10 Thus the reduplicative structure is improved, as the outcome is no longer a singleton consonant, but a whole syllable, on the assumption that prosodic reduplication is better than the doubling of bare segments.

On the borderline of reduplication

383

While full reduplication as well as final and initial consonant doubling are definitely not productive (yet) and are sometimes not derived from an existing root, it seems justified to assume that new reduplication rules have come into being. Many of the CVCCVC verbs show a sonorant, very frequently Irl or the semivowel /w/, which is the typical default consonant in Arabic word formation, in the position of C2. The scenario for the development of a new reduplication rule might thus be as follows: Degemination and the deconstruction of a fuller reduplicated form have produced novel forms with apparently doubled Q or C 3 which might be illustrated by the following examples from MA and EA, which both have an ancestor in C A f a t t . f a t t a t 'to crumble bread'. 'fattat > fartat 'to crumble bread, to disperse' (MA) fatfit > farfit 'to crumble bread' (EA) These forms now serve as the basis for analogical formations that aim at replicating the novel pattern. Note that not every surface form exhibiting a doubled consonant can be derived in this way. Such forms might thus be considered as analogical formations, especially when they cannot be traced back to any original root. There are dozens of verbs, like EA dardif, f a r f l f , dandif etc. and MA bsrgsg, tentst, ζ on hi, gu393 etc. that might have been partly formed by analogy." Following Ratcliffe's (2001) proposal that analogy involves the creation of a rule on the basis of surface patterns, I assume that these analogical formations in turn lead to a rule that is applied to triliteral roots, to derive verbs like lahlib 'to cause a burning sensation' from lahab 'flame'. Contrary to initial doubling, final doubling, of course, has a history in Semitic, in CA in the formation of the IXth and XIth and XIVth stem, but the CVCCVC-forms of that type only appear in NeoArabic, esp. in the Maghrebinian varieties. From the examination of such forms in Arabic, it becomes clear that the only productive reduplication rules in CVCCVC-verbs in Neo-Arabic are gemination and perhaps full reduplication of biconsonantal roots, the other two kinds of reduplication still having a marginal status in the grammar. In order to evaluate the different kinds of reduplication in Arabic verb forms, we should consider the principles of reduplication that have been assumed in section 2.3 and are replicated here for convenience. If we assume that the formal procedure of reduplication originates in the repetition of phonological material as in babbling for example, which may be utilized to denote grammatical categories or notions, the best type of reduplication seems to be one that doubles whole structures as they are. This means that

384

Dina El Zarka

full reduplication is better than partial reduplication and reduplication of prosodic units with the minimal size of a syllable is better than the doubling of bare segments. Within the realm of segmental doubling, the reduplicant, in this case the copy of the segment, should be in the same syllabic position as in the base, i.e. it should alliterate or rhyme with the segment of the base. Under this assumption, Arabic verb forms can be graded on a scale from "best" to "worst" type of reduplication: katkat > katkab > katbab > kattab The segments in the katkat-form alliterate and rhyme; the output shows two identical syllables. In katkab the two syllables have the same onset-C, so at least alliteration is realized, whereas in the last two forms the copied consonants neither alliterate, nor do they rhyme. The problem with geminates is that they are articulated and also perceived as one gesture, which makes the doubling more opaque and also makes them prone to degemination. It has been observed that reduplicative constructions may be subject to phonetic erosion. Thus, full reduplication may develop into partial reduplication and syllable reduplication may eventually result in gemination after the syncope of a vowel, as for example in Trukese (Goodenough 1963). This is what the strong grammaticalization hypothesis of reduplication predicts. As has been shown, the case of Arabic does not sustain this hypothesis, as geminates do not originate in fuller forms of reduplication, but on the contrary, the geminated forms are definitely older than the other reduplicated forms. In addition to the frequently observed deterioration or loss of reduplication, there are cases attested cross-linguistically where reduplicated structures are improved (cf. Hurch and Mattes, to appear). Of course, these are not very common and the main process is the one of deconstruction. In Arabic CVCCVC verbs, both processes, a) deconstruction and b) structural improvement are attested: a)

full reduplication > partial reduplication partial reduplication > zero-reduplication

b)

partial reduplication > full reduplication

Examples for (a) have been cited above in 3.1.3, such as the emergence of a sonorant in the coda of the first syllable of fully reduplicated forms like kabkab > karkab, thereby giving rise to the novel type of initial reduplica-

On the borderline of reduplication

385

tion and the reduction of geminated triliterals as in faqqa? > farqa?, which is an example for the complete loss of reduplication and the emergence of novel quadriliteral forms. Examples for b) are all instances of fully reduplicated biconsonantal roots that are an innovation, compared to the older forms with doubled C2 as in fattit > fatfit or laffif > laflif (cf. 3.1.2, table 2). I do not assume, though, that these novel forms are directly derived from the geminated pattern, but rather that they replace it.

5.

Formal analysis

Having established the semantic, diachronic and structural relationship of all different CVCCVC forms, some kind of formal analysis seems to be called for. Formalizing linguistic rules is always a highly hypothetical and tentative thing to do, as the history of modern linguistics amply shows. But, on the other hand, it may lead us to gain deeper insight into the actual procedures that go on in our minds in the process of linguistic production, as well as into the mechanisms of language change. The major problem is that the wish to make the model work will often make the data obey a certain system, instead of finding a system that explains the data. Among other issues, linguistic debates of the past decade have centered around the (prosodic) shape of the reduplicant and its base. Starting with Marantz (1982), reduplicants have been defined phonologically as segmental strings, or later in prosodic terms as exemplifying a certain prosodic shape (McCarthy and Prince 1986 et seq.). Clearly, the case of Arabic provides no evidence for the latter assumption. Gafos (1995) proposes to analyze consonant reduplication as "a-templatic" reduplication, i.e. not satisfying a prespecified prosodic template, relying on data from the Austronesian language Temiar, the formation of Hebrew denominative verbs and Arabic root formation. He also proposes that the distinction between templatic and a-templatic reduplication is due to the difference between concatenative and non-concatenative morphology. Broselow and McCarthy (1983) analyzed reduplication in Arabic katkab and katkat forms as essentially the same process of infixation to a prespecified C-slot. However, they did not address £a?6a&-reduplication and gemination in the kattab-pattern is explicitly excluded (p. 76). In McCarthy (1981, 1982) the geminate was ascribed to the addition of a segment and the spreading of phonetic information, and in McCarthy (1992) this approach was refined and the augmentation of the template was ascribed to the addition of a mora within the theory of prosodic circumscription.

386

Dina El Zarka

As I have tried to show in the previous section, the different kinds of doubling in the disyllabic Arabic verb should rather be viewed as related and thus the mechanism of their derivation should also be subject to a unified treatment. As outlined above, I assume the CVCCVC pattern to be an augmentation of the prosodic shape, which is achieved by the addition of a mora to the simplex verbal pattern CVCVC (LL),' 2 thereby yielding a sequence (HL). The actual segmental shape of the pattern is then arrived at by copying a consonant: C], C2 or C3. The locus of the reduplicant is the onset of the second syllable. The most recent development in generative treatments of reduplication has distinguished between prosodic and morphological bases of reduplication. As reduplication in verb formation is a morphological operation, it seems only natural that it should take as its base a morphological entity, such as the root. There is cross-Semitic evidence (Unseth 2003) as well as evidence from syntactic reduplication in Arabic (Maas, this volume) that it is the root consonants that are doubled, and, as the data show, every consonant can in principle be doubled. The linguistic output, however, is subject to prosodic constraints. As McCarthy (1982) already noted, Arabic katkab-reduplication is paralleled in Temiar. In fact, the following Temiar examples, often cited in the literature, are the mirror image of the Arabic verbs. 13

Temiar: triconsonantal root

biconsonantal root

perfective

s.log

k5w

continuative

sg.bg

kw.kSw

triconsonantal root

biconsonantal root

perfective

farah

laff

pluractional (perf.)

farfah

laflif

Arabic:

In Broselow and McCarthy's (1983) analysis the difference between the Arabic and the Temiar examples lies in the "direction of association of the phonemic melody copy with the infix" (p. 38). Within Optimality Theory,

On the borderline of reduplication

387

Gafos (1995, 1998) translates the analysis of the Temiar data to prefixation or alignment to a stressed syllable or the prosodic head of the prosodic word. The parallelism between the Temiar and the Arabic examples is definitely striking. Thus, we could also venture to analyze reduplication in the Arabic forms in terms of "suffixation" or of "alignment to the right" of the stressed syllable which is the head of the foot, as Arabic disyllabic verbs have trochaic stress. As we want to include all kinds of doubling in the CVCCVC pattern of Arabic verbs into our analysis, this clearly demands that 1) copying must take place from left to right or from right to left; as the target is in infixal position, this option is not to be excluded a priori.

root °frh° °k-s-r° °[-m-l°

simplex farah kasar Jamal

CVCCVC farfah kassar Jamlal

direction of copying left-to-right left-to-right right-to-left

2) The base to which the reduplicant is aligned and the input for copying are different, the former being a prosodic constituent, viz. a stressed syllable, the latter a morphological one, viz. the root. Again, 1 take this to be in line with the assumption that reduplication is a morphological procedure that copies part of a morphological entity, but that its output is subject to prosodic and other phonological requirements or constraints. There are other principles that govern the outcome of reduplication, whether they are strict requirements which cannot be violated, or just preferences, like the aforementioned alliteration and rhyme that can be violated. I will not dwell on these however, as they are not the subject of this paper. Whatever the correct analysis may be, the major insight from the parallelism with the Temiar facts seems to be the difference in the role of the prosodic template within morphology, as already pointed out by Gafos. While in concatenative morphology it is the shape of the reduplicative "affix" that is prosodically determined, in non-concatenative morphology it is the prosodic shape of the whole structure that is fixed. Consequently, in Temiar as in Arabic, whether one or two consonants are doubled depends on the number of root consonants, because the requirement that must be met in the first place is the shape of the disyllabic template.

388

6.

Dina El Zarka

Conclusion

In this article I have tried to establish the relationship of different consonant doubling phenomena, including gemination, in Arabic verbal morphology. Furthermore, I have provided some new data of Arabic CVCCVC-verbs that show the emergence of a pattern of full reduplication, not just for a lexical class of onomatopoetics as in Old Arabic, but also as a derivational class, which seems to exist beside the ancient stem II (D-stem) and is even partly replacing it. If the proposed analysis is right, this will answer the questions asked in the introductory paragraph of this paper: (i)

Is reduplication of bare segments only a phonological operation or does it bring about semantic change and (ii) does reduplication necessarily obey prosodic requirements or can the doubling of bare segments be counted an instance of reduplication? The analysis of the data has shown that the doubling of consonants in Arabic is clearly a morphological issue, which changes the derivational class and the semantics of the verb. Considering the structure of the root, which only consists of consonants, it seems only natural that bare consonants are doubled. Moreover, if we compare doubling with affixation, we find that an affix in Arabic typically has the size of just one segment, e.g., two derivational affixes in Arabic verb formation are °t° and °n°. Furthermore, the doubling of bare segments is a legitimate form of morphological reduplication, connected to the morphological type of Semitic, in which the consonantal root is the basic lexical morpheme. Segmental doubling, though, is a rather bad instance of reduplication as far as constructional iconicity is concerned, as it is harder to perceive and process than fuller reduplications that abide by the extra-grammatical principle of alliteration and rhyme. (iii) Is gemination in Arabic to be considered as reduplication or spreading? It has been shown that the geminate clearly is an instance of two identical consonants. Thus gemination in Arabic verbs should be viewed as reduplication. On the other hand, as has happened in some Semitic languages, gemination can easily become fossilized and lose its morphological function. As geminated consonants are phonetically one, it is only natural that with time they should develop into singleton consonants like in Hebrew

On the borderline of reduplication

3 89

and thus become prone to consonant mutation phenomena. What started out as morphological doubling can become consonant mutation, following the grammaticalization path of phonetic and semantic erosion. (iv) Is the strong grammaticalization hypothesis (Niepokuj 1997: 63) that predicts a grammaticalization path along the following line: full reduplication > partial reduplication > [ . . . ] > gemination correct? The case of Semitic provides clear evidence against a strong grammaticalization hypothesis, as it becomes evident that partial reduplication does not necessarily belong to a later stage of earlier fuller or total reduplication. It rather seems that different reduplications can coexist to convey different meanings (cf. also Haugen, this volume, Hurch and Mattes, to appear). On the contrary, as in the cases of the Ethiopian Semitic languages, Modern Arabic and Hebrew show that fuller reduplication may develop in a later stage of the language. Gemination as consonantal doubling in Arabic word formation has to be viewed as a basic derivational process as it definitely is not the outcome of a phonological truncation process like vowel deletion and is therefore not the endpoint of grammaticalization.

Notes 1.

2.

3.

4.

I am greatly indebted to Robert Ratcliffe for his valuable comments and constructive criticism on a prior version of this paper. Special thanks to Bernhard Hurch for reading the manuscript and many helpful comments and to Veronika Mattes for discussions. The examples in the text come from the following Arabic varieties: Classical Arabic (abbreviated as CA), Modern Standard Arabic (MSA), Egyptian Arabic (EA), Levantine Arabic (LA) and Moroccan Arabic (MA). For a similar view cf. Rose (2003), who also gives a short introduction into the debate. Evidence for the root and the template comes from language games (McCarthy 1982) and from psycholinguistic experiments. The sources of the data are Cowell (1964) for LA, Imouzaz (2002) and my Moroccan informant (Casablanca) for MA and Hinds and Badawi (1986) and my Egyptian informants for EA. The /a/ denotes the inflectional ending of the 3.m.sg. which has been lost in Neo-Arabic and is also dropped in pausal form. It has to be stated here because the verb is embedded in a whole sentence.

390 5. 6.

7.

8.

9.

10.

11. 12.

13.

Dina El Zarka Scanning the dictionaries of Wahrmund, Lane, Freytag and Blächere, Prochäzka found 417 reduplicated roots (p. 43). For reasons of space I will not go into details here. The interested reader is referred to Brockelmann 1908, Kienast 2001, Lipmski 2001 where further references are cited. Robert Ratcliffe (personal communication) has suggested that causatives involve an increase in arguments, thus this construction might also be viewed as iconic. The bulk of linguistic studies on reduplication/consonant spreading in Arabic (starting with McCarthy 1986) is concerned with the formation of the basic stem from a biconsonantal root (the so-called geminated roots of traditional grammar). Despite the structural relationship between these forms and verbs with final doubling, I have not included them in my argumentation, as I consider this subject to be a matter of diachronic root formation and not of synchronic morphological derivation; there is no difference in meaning between a root °mr° and °mrr°, the expansion of the root is rather a matter of achieving the ideal triconsonantal root. In Neo-Arabic varieties the geminates in these forms have become fossilized: marr 'he passed' and marre:t Ί passed', ihmarre:t Ί got red'. Stem IX is not a derivational class but a lexical one with only a few tokens. In MA where consonant clusters are more freely allowed, the forms with final reduplication show a geminate in some inflections: bargsg + u (3.pi.) > barggu 'they had an eye on'. Unseth (2003: 268) cites some word forms from Ethio-Semitic languages like Tigre gälbä > ?ägläbläba and GeVez rämsäsä > ?ärmäsmäsä that provide evidence for the view that it is essentially the consonants that are reduplicated. isntat for example, may go back to Vatt 'to reiterate questions, blame s.o.' as the base for a derivation: °itt°: *fott3t > Vantat 'se rebeller.' Note that a heavy final syllable (CV: or CVC) is prosodically light due to its peripheral position and thus does not receive stress resulting in a trochaic stress pattern for such word forms (cf. El Zarka 1999). Reduplicants are printed bold-face and a dot signifies a syllable boundary.

References Angoujard, Jean Pierre 1988 Gemination et redoublement. Langues orientales anciennes - philologie et linguistique 1: 1-15. Bat-El, Outi 1994 Stem modification and cluster transfer in Modern Hebrew. Natural Language and Linguistic Theory 12: 571-596.

On the borderline of reduplication

391

Brockelmann, Carl 1908 Grundriss der Vergleichenden Grammatik der Semitischen Sprachen, Bd. 1. Laut- und Formenlehre, Berlin. 1927 Semitische Reimwortbildungen. Zeitschrift für Semitistik und verwandte Gebiete 5: 6-38. Leipzig: Deutsche Morgenländ. Gesellschaft. Broselow, Ellen and John McCarthy 1983 A theory of internal reduplication. The Linguistic Review 3: 25-88. Cowell, Mark W. 1964 A Reference Grammar of Syrian Arabic. Georgetown University Press, Washington, D.C. Cusic, David D. 1981 Verbal plurality and aspect. Ph.D. dissertation, Stanford University. Dressier, Wolfgang U. 1968 Studien zur verbalen Pluralität. Wien: Verlag der Österreichischen Akademie der Wissenschaften. 1990 The cognitive perspective of "naturalist" linguistic models. Cognitive Linguistics 1: 75-98. 1996 Principles of naturalness in phonology and across components. In Natural Phonology: The State of the Art, Bernhard Hurch and Richard Rhodes (eds.), 41-52. Berlin/New York: Mouton De Gruyter. El Zarka, Dina 1999 Der Akzent im kairinischen Arabisch - eine natürlichkeitstheoretische Beschreibung. Grazer Linguistische Studien 52: 53-72. Gafos, Adamantios 1995 On the proper characterization of "non-concatenative" languages. ms., Cognitive Science Department, The Johns Hopkins University, Baltimore, Md. Rutgers Optimality Archive ROA-106, http:// ruccs.rutgers.edu/roa.html. 1998 Eliminating long-distance consonantal spreading. Natural Language & Linguistic Theory 16: 223-272. Goodenough, Ward H. 1963 The long or double consonants of Trukese. In Proceedings of the Ninth Pacific Science Congress, 77-80. Bangkok. Greenberg, John 1950 The patterning of root morphemes in Semitic. Word 6: 162-181. 1991 The Semitic "Intensive" as verbal plurality. In Semitic Studies: in Honour of Wolf Leslau on the Occasion of his 85lh Birthday, Alan S. Kaye (ed.), 576-587. Wiesbaden: Harrassowitz. Hendricks, Sean 1999 Reduplication without Templatic Constraints. A study in BareConsonant Reduplication. Ph.D. diss., University of Arizona. ROA 511-0402

392

Dina El Zarka

Hinds, Martin and El-Said Badawi 1986 Dictionary of Egyptian Arabic. Arabic-English. Beirut: Librairie du Liban. Höfner, Maria 1951 Probleme der Verbalstammbildung im Tigre. Zeitschrift der Deutschen Morgenländischen Gesellschaft 101: 89-106. Hurch, Bernhard and Veronika Mattes 2004 Über die Entstehung von partieller Reduplikation. In Sprache und Natürlichkeit. Gedenkband für Willi Mayerthaler, Gertraud FenkOczlon and Christian Winkler (eds.), 137-156. Tübingen: Narr. Ibrahim, Muhammad H. 1982 Radical reduplication in Arabic. Zeitschrift für Arabische Linguistik 9: 82-87. Imouzaz, Said 2002 Interaction des contraintes dans la morphologie non-gabaritique de l'Arabe Marocain de Casablanca. Temoignages pour la Thiorie de l'Optamalite. These de Doctorat. Universite Hassan II. Maroc. Kienast, Burkhart 2001 Historische semitische Sprachwissenschaft. Wiesbaden: Harrassowitz. Lipmski, Edward 2001 Semitic Languages: Outline of a Comparative Grammar. (Orientalia Lovaniensia Analecta 80), 2 nd ed. Leuven/Paris/Sterling: Peeters. Marantz, Alec 1982 Re reduplication. Linguistic Inquiry 13: 435^182. McCarthy, John 1979 Formal problems in Semitic phonology and morphology. Ph.D. diss. MIT. 1981 A prosodic theory of nonconcatenative morphology. Linguistic Inquiry 12: 373-417. 1982 Prosodic templates, morphemic templates, and morphemic tiers. In The structure of phonological representations I, Harry van der Hulst and Nancy Smith (eds.), 191-223. Dordrecht: Foris. 1986 OCP Effects: Gemmation and Antigemination. Linguistic Inquiry ΜI2\ 207-263. 1992 Template Form in Prosodic Morphology. In Papers from the 3rd Annual Meeting of the Formal Linguistic Society of Mid-America, Laurel S. Stvan (ed.), 187-218. McCarthy, John and Alan Prince 1986 Prosodic Morphology, ms. University of Massachusetts and Brandeis University.

On the borderline of reduplication

393

Moravcsik, Edith 1978 Reduplicative constructions. In Universals of Human Language. Vol. 3, Word Structure, Joseph Greenberg (ed.), 297-333. Stanford: Stanford University Press. Moscati, Sabatino, Anton Spitaler, Edward Ullendorff and Wolfram Von Soden 1980 An Introduction to the Comparative Grammar of the Semitic Languages. Phonology and Morphology. 3rd Printing. Wiesbaden: Harrassowitz. Niepokuj, Mary 1997 The Development of Verbal Reduplication in Indo-European. (Journal of Indoeuropean Studies Monograph 24). Washington: Institute for the Study of Man. Plank, Frans 1981 Morphologische (Ir-)Regidaritäten: Aspekte der Wortstrukturtheorie. Tübingen: Narr. Pott, August F. 1862 Doppelung (Reduplikation, Gemination) als eines der wichtigsten Bildungsmittel der Sprache, beleuchtet aus Sprachen aller Welttheile. Lemgo/Detmold: Meyer. Prochäzka, Stephan 1995 Semantische Funktionen der reduplizierten Wurzeln im Arabischen. Archiv Orientalni 63: 39-70. Ratcliffe, Robert 1998 The "Broken" Plural Problem in Arabic and Comparative Semitic: Allomorphy and Analogy in Non-concatenative Morphology. Amsterdam: Bemjamins. 2001 Analogy in Semitic morphology: Where do new roots and new patterns come from? In New Data and New Methods in Afroasiatic Linguistics: Robert Hetzron in Memoria, Andrzej Zaborski (ed.), 153162. Wiesbaden: Harrasowitz. Rose, Sharon 2000 Rethinking geminates, long-distance geminates, and the OCP. Linguistic Inquiry 31: 85-122. 2003 The formation of Ethiopian Semitic internal reduplication. In Processing and Acquisition of Semitic Morphology, Joseph Shimron (ed.), 79-97. Amsterdam/Philadelphia: John Benjamins. forthc. Triple Take: Tigre and the case of internal reduplication. San Diego Linguistic Papers 1. Sapir, Edward 1921 Language. New York: Harcourt, Brace, and World.

394

Dina El Zarka

Unseth, Pete 2003 Bi-consonantal reduplication in Semitic. In Selected ComparativeHistorical Afrasian Linguistic Studies in Memory of Igor M. Diakonojf, M. Lionel Bender (ed.), 257-273. Munich: Lincom Europa. Wehr, Hans 1977 Arabisches Wörterbuch für die Schriftsprache der Gegenwart, 4. unveränderte Auflage, Beirut: Librairie du Liban. Wiltshire, Caroline and Alec Marantz 2000 Reduplication. In Morphologie/Morphology. An International Handbook on Inflection and Word-Formation. Vol.1, Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), 557-567. Berlin/New York: Walter de Gruyter. Wright, W. 1974 Reprint. A Grammar of the Arabic Language. 3 rd ed. Beirut: Librairie du Liban. Original edition 1896. Yip, M. 1999 Reduplication as alliteration and rhyme. Glot International 4 (8): 1-7.

Syntactic reduplication in Arabic Utz Maas

1. Iteration and reduplication - the case of grammaticalization. Some introductory comments In current discussions, reduplication is generally treated as a device of word formation, i.e., as a special form of derivational patterning (cf. e.g. Wiltshire and Marantz 2000; Raimy 2000). In this paper I will take a more general perspective, which can already be found in older typological work (e.g. Pott 1862): Reduplication is seen as a formal linguistic device that can be used at all levels of linguistic structure. This theoretical move shifts the leading question to one of grammaticalization: To what extent is reduplication grammaticalized 2 . To get a better grasp of the problem, we need a pretheoretical (or pre-grammatical) term: In this sense I will speak of doubling or in the case of more than two forms of iteration. Investigating this phenomenon produces a problem similar to optical illusions: The way you look at it determines what you will see. Doubled forms can be the result of a lack of structural differentiation, as e.g. in early child language and in certain types of aphasic speech, or it can be the result of structure building, i.e. as the effort to keep things indentical through repetition. Two perspectives on doubling can thus be distinguished: I.

Doubling can be seen as a reflex of inertia in linguistic activity,

II.

It can also be seen as the result of a special type of structuring, thus as learned behavior in a linguistic community, as part of the language system.

Conflation of these two different perspectives prevails not only in lay thinking about language, where it is considered stereotypical of "primitive" languages. Even grammatical handbooks, especially when addressing a larger public, hawk this stereotype as well3. It is an aim of the present paper to differentiate between perspectives I and II which must be understood as dual aspects of linguistic behavior and thus potentially present in any speech act or language. We can trace perspective I at all levels of language:

396

Utz Maas

from echoed prosodies in conversation up to lexical repetitions. The need to differentiate between perspectives I and II is especially important in research on early language development: Doubling is among the earliest devices a child makes use of when trying to impose structure on its utterances, cf. Leroy and Morgenstern (this volume), using the more or less biological constraints of its linguistic activities as resources for symbolic behavior, successively imposing more differentiated structures upon its utterances in pace with its mastering of the structures of adult language. But this ambivalent relation to doubling is also present in adult behavior: Every speaker (or writer) can make an effort to use iteration as well as an effort to avoid it. This must be distinguished from grammaticalized reduplicating patterns in the structure of a language which cannot be avoided (i.e. which are not mere stylistic options). This distinction can already be found in early research: Hermann Reckendorf, a pioneer of reduplication research 4 , distinguished bound reduplication (gebundene Paronomasie) from stylistically free reduplication (Reckendorf 1909). Bound reduplication pertains to grammaticalization. This is the subject of the present paper in which I will present some data from a linguistic family that is notorious for its extensive use of reduplication: Arabic 5 . On a very general level, perspective I implies the idea that language is articulated, i.e., formally differentiated in its elements. Thus where iteration is not due to imperfect or rudimentary linguistic knowledge, it is a marked form of expression (= perspective II). Markedness conveys a particular interpretation. Examples of this can probably be found in all languages, cf. the German example (l) 6 : (1)

Es kommt selten vor, dass sich der hessische Ministerpräsident windet und windet und windet, wenn ihn jemand nach seiner Meinung fragt. 'It seldom happens that the Prime Minister of Hessia squirms and squirms and squirms when someone asks him for his opinion'

Expressions like these can be more or less conventional or even lexicalized (as idioms). In non-lexicalized cases, the interpretation is a function of the basic meaning of the iterated elements, the markedness of the iteration adding, e.g., an element of gradation, cf. the German examples in (2), heard in injunctions, that can probably be replicated in most (all?) languages: (2)

langsam, langsam! 'slowly, slowly!' (i.e. calm down!) schnell, schnell! 'quickly, quickly!' (i.e. hurry up !)

Syntactic reduplication in Arabic

397

On the formal side, doubling or iteration can be holistic or partial (analytic). Holistic iteration can be achieved by the simple repetition of utterances or their parts, e.g. words as in (1) and (2). In distinction to this, grammaticalized devices are to be expected to be analytic. This can be exploited in "poetic" uses of language, e.g. in the comic series the SMURFs (in German: Schlumpf), very popular some years ago. The Smurfs had a special Smurf language, with fully developed grammatical devices but with a lexical dummy smurf that could be used throughout. Thus, in the smurf language it would have been possible to build a sentence like: (3)

I smurfed my smurfy smurf to a smurf

Besides the emphasized functional elements, there are of course covert grammatical devices such as e.g. word order, that show English as the matrix language. Iteration in (3) is analytic: The word forms in (3) are differentiated by grammatical formatives, the iterated element smurf is a stem. A language with sentences like (3) would be undifferentiated (unarticulated) only at the level of the lexicon, not at the level of word-forms. In fact, the smurfs were rather parsimonious in using this kind of reduced differentiation. In the comics we do not find sentences like (3) but rather like (4): (4a) (4b)

Lazy Smurf, have you smurfed that play for our village fair? And look at Smurfette! She 's much smurfer than that!

Smurf-forms are marked expressions in an otherwise lexically differentiated context 7 . And every smurf-form can be substituted by a fully differentiated non-smurf form. Thus we can hypothesize that elements of low differentiation fulfill a special function in full-fledged languages, both in the Smurf language as well as in "natural" languages. The grammaticalization of doubling (i.e. reduplication) implies the definition of the domains where it applies, in contrast to iteration which is unbounded and can be exploited by all kinds of parallelisms in texts, especially in poetic language. The domain of iteration in the Smurf-language is syntax (the sentence) and the form of iteration is analytic: Lexical stems are repeated, augmentend by grammatical formatives that determine the syntactic function of the word forms. Analytic reduplication presupposes the parsing of forms, which can be done on a morphological basis, e.g. by parsing word forms into stems and affixes as in (3). Depending on the language type, the categories of morphological parsing can be more fine grained, e.g. roots can be iterated as in Semitic languages (see below), and there can be

398

UtzMaas

phonological filters as well. The generalized pattern of this type of reduplication isolates word-internal elements that are lexically anchored. Thus the structure of a sentence like (3) can be graphically represented as (5), where Greek letters represent (grammatical) affixes and Roman letters the stems: (5)

#

Reduplication through the repetition of word-internal lexical elements in a sentence (reduplicands are represented by capitals): Α



Β



C



D



Ε



#

Here, A, Β, C, D and Ε are derived from the same stem, while the affixes indicate the function of these words in the construction. The mirror of (5) iterates the grammatical formatives (operating again within a syntactic domain) and is graphically represented in (6). (6) #

Grammatical reduplication: Α-

α

Β-

β

C-

γ

D-

δ

Ε-

ε

#

Here, Α, Β, C, D and Ε derive from different stems while the affixes are identical (in function, at least). It is evident that (6) represents what is usually called agreement, cf. the Latin example (7): (7)

tenebr-ae altissim-ae obor-t-ae darkness-N.PL.F profoundest-N.PL.F ensue-PCP-N.PL.F 'deepest darkness fell'

sunt be:3.PL

Following the classification of Pott (1862), I will call the types of reduplication found in (5) and (6) syntactic reduplication. The main subject of this paper will be the structure (5). The domain of reduplication can be smaller than the sentence. Grammaticalization presupposes a grammatical constituent as its domain: i.e., a construction. The limiting case of construction is the (morphologically complex) word. Thus in the limiting case of reduplication, form elements are reduplicated word-internally. In languages with morphological affixes, this leads to structures like (8):

Syntactic reduplication in Arabic (8) #

399

Word-internal reduplication: -a

As the delimitation between constructions and words is rather complicated from a typological point of view, we can expect these complications in reduplication research as well. This will become evident in the following, when cases of syntactic reduplication are analyzed in which the domain can perhaps be understood as "multi-worded" words 8 . Word-internal reduplication is the dominant subject of current research in reduplication, in general presupposing the word as the domain of reduplication, but without bothering much about defining this entity. An exception is Moravcsik (1978), who uses a broad concept of reduplication, one that provides for the analysis of syntactic reduplication as well. However, in her typological survey, she only takes word-internal reduplication into account. Recent work simply presupposes word-internal reduplication as the domain of reduplication, e.g. Marantz (1982), and Raimy (2000), where the focus is on the phonological constraints. I will take the broader perspective, and will use Moravcsik's reduplicative construction as a cover term for the different types of reduplication shown in (5), (6) and (8), analyzing the special kind of structural unity it imposes on its constituents. Reduplication as an analytical device presupposes some kind of filter to define the (partial) structure that has to be reduplicated. The filtering patterns can be based in phonology (prosodic or syllabic templates) or can be purely formal (morphological, without alignment to phonological structures). These patterns are the subject of most ongoing research, which aims at identifying universal patterns (e.g. the "emergence of the unmarked" in reduplicands). The focus of this paper is complementary to this approach: I will look at reduplication devices (analytic iteration) as the germ of grammaticalized structures, where the reduplicand modifies the base, as in (8). This makes it necessary to delimitate reduplication from holistic doubling, which borders on stylistics, as well as from exclusively formal iteration without functional differentiation, be it in lexical structure or as a kind of stylistic play with form, as e.g. in alliteration, rhyme, etc9. In the following I will focus on syntactic reduplication. This has traditionally been the subject of rhetorical studies, where it is discussed as paronomasia. Paronomasias can produce tautological expressions, coming close to the Smurf language, but even then they can acquire meaning, as in Gertrude Stein's famous a rose is a rose is a rose. But in a more trivial con-

400

Utz Maas

text they run the risk of appearing rather nonsensical, as e.g. the watery water, the hiker hikes, he screamed a scream, etc. Usually, they will only make sense if further modified, for example: (9)

Types of syntactic reduplication subject - predicate: the hiker hiked in the mountains noun - attribute: the very watery water (??) verb — object: he screamed a loud scream

Especially problematic is the paronomastic attribute (9b); the other constructions are more acceptable, where the paronomastic element serves to support a further meaningful specification. Despite this, there is a long normative Western tradition which frowns upon paronomasias, already present in classical times 10 . This makes for an interesting contrast to Semitic traditions, where texts abound in them: They are favored not only in poetic works (including the Qor?an), but are current in colloquial language as well (cf. Reckendorf 1909 for documentation from literary as well as nonliterary texts). These different stylistic preferences might be a hint at differences in the "underlying" linguistic type: Semitic languages are obviously more prone to this device than Indo-European languages. This will be explored in the following, focusing on reduplication within the verb phrase (cf. 9c) and mentioning other types only cursorily.

2.

Iteration and reduplication in Old (= Classical) Arabic

In this paper, I will use the term Arabic in the generic sense of a phylum, thus including Old Arabic as well as neo-Arabic varieties. When the reference is more specific, I will use a more specific term. It is important to differentiate between Old Arabic in the diachronic sense, for which I take Classical Arabic as a grammatically standardized representative, and (Modern) Standard Arabic. As diachronic aspects are of interest here, I will contrast Old Arabic, quoting form reference works for the Classical texts, with data from two neo-Arabic varieties, Maltese and Moroccan Arabic. At first glance, the structure of Arabic is characterized by a high degree of grammatical differentiation, with a fusional morphology that has been retained, or more precisely reconstructed, in neo-Arabic varieties despite the very far-going changes in phonology, such as apocope of inflectional suffixes, loss of quantity opposition in the vowel system, and the complete restructuring of syllabification (especially in Western varieties). An exam-

Syntactic reduplication in Arabic

401

pie of syntactic reduplication in Classical Arabic (or Modern Standard Arabic) is (10):

(10)

katab-a

l-ka:tib-u

write:PF-3SM DEF-writer-NS 'the writer wrote the book'

l-kita:b-a DEF-book-AS

(10) is different from (9) as well as from the Smurf examples in (4) as there is no perfect match between the base and the reduplicated elements: Reduplication is only partial as far as the segmental/concatenative structure is concerned. Thus we could call the reduplication dense in cases like (9), where the reduplicands are complete strings in the form of the base, and porous in cases like (10), where the identifiable reduplicand is only matched by a porous structure in the base, in (10) the sequence of consonants k-t-b. But a classification such as dense vs. porous would miss important generalizations of grammaticalization governing reduplication in Arabic. (10) can be represented in a more explicit (transparent) way as (11):

(11) kataba l-ka:tibu l-kita:ba

= ktb {V-PF.3SM} + ktb {N-Ag.NS} +

ktb {N-Res.AS} 'the writer wrote the book' (11) shows the invariable element °ktb° with internal grammatical markers (indicated by {}): The finite verb, with the functional marking of the predicate, the main actant, articulated as "agentive noun" (N-Ag) in the subject case, and the secondary actant, articulated as "resultative noun" (N-Res) in the object case. This construction is bound to the special structure of Semitic languages, which split the stem into different melodies. Arabic has been the paradigmatic case in recent morphological work operating on a multi-layered analysis (where the layers must be defined independently), in which dense structures {tiers) at the different layers are identified. Autosegmental approaches have generalized this model since McCarthy (1982), and most recent work in reduplication (word-internal reduplication, cf. (8)) has generally taken this model as its guide-line (since Marantz (1982)). As this multi-layered analysis of Arabic is well-known, it will be sufficient to illustrate it with one example from the verb form kataba "(he) wrote". The structure is presented in a grid, separating the different dimensions11:

402

Utz Maas

1. 2.

Word-level, simplified as segmental (time) slots, melodies, differentiated for consonantal and vocalic elements (defined by syllabificational potentials [potential nuclei vs. margins]) root-level: lexical invariants, traditionally called "radicals", stem-level: augmented root-structure, modified by inflectional affixes, to be differentiated between a. "word formation": semantic modification of the root (stem-W), b. grammatical specification (stem-G) inflectional level: prefixes and / or suffixes

3. 4.

5.

Levels (1) and (2) define the phonological structure, level (3) defines the lexical invariant, and the function of the chosen form is indicated at levels (4) and (5). Thus levels (3) to (5) can be seen as increasing specifications of the word form: Level (3), the root skeleton, is the minimal specification, level (5) is the maximal specification: (12)

word-form (verbal) kataba (six time slots)

1. word

X C

2. melodies

X

V

3. root

X

"he has written" =

X

X

X X

R, = k

{x-x-x-x-x-x}

X

X

X X

R2= t

X R3 = b

{writing} 4. a stem-W (void)'2 4.b. stem-G

V, = a

V2 = a

{ACT, PF} 5. inflection {3.S.M.}

V3 = a

Syntactic reduplication in Arabic

403

Examples such as (11) show the partial reduplication of words at the syntactic level. The word forms are morphologically differentiated for syntactic functions. Holistic reduplication is possible as well, where the reduplicated forms fill the syntactic slots of words, cf. (13)

holistic reduplication in Classical Arabic13:

a.

b.

c.

ma/a:

fuwaj:a(tan)

fuwaj:a(tan)

go:PF.3.S.M. 'he went slowly'

little (A.S.-DEF)

little (A.S.-DEF)

?intamr-a

sanat-an

sanat-an

look.out:PF-3.S.M year-A.S.-DEF 'he looked out year after year'

year-A.S.-DEF

?istaqra:hum

haj:-an

haj:-an

exam ine: PF-3. S. Μ tribe-A.S.-DEF 'he examined them tribe for tribe'

tribe-A.S.-DEF

Reduplication here is a means of semantic modification that would have an equivalent in word formation in many other languages (Arabic is especially poor in terms of word formation devices). Thus cases like (13) represent borderline cases between syntactic and word-internal reduplication, since a constructional meaning can be identified. The modification of the meaning is usually quantifying. Thus the reduplicative means is in a certain sense "iconic"14, often corresponding to determiners (quantifiers, adjectives etc.) in other languages. It is often expressed through equivalent syndetic (coordinated) constructions, which could be used in other languages as well, cf. the examples of syndetic reduplication in Old Arabic in (14):

(14) a.

na.da:

radjul-un

wa

call:PF.3.S.M man-N.S.-DEF 'every individual man called'

b.

hazin-at

and

radjul-un man-N.S.-DEF

Talaj-ha: ?akdara

become.grieved:PF-3.S.F to-3.S.F 'she always becomes grieved'

more

fa Takdara and more

Apparently the kind of reduplication in (13) and (14) is a stylistic means, and as such possible in other languages as well (such as English), which are much less tolerant of asyndetic expressions, preferring syndetic expressions instead. In the following, this kind of reduplication will not be further ana-

404

Utz Maas

lyzed, and I will focus on cases of analytic syntactic reduplication. But as word-internal reduplication is often quoted as characteristic of Arabic15, these structures will also be discussed briefly. In the lexicon, the doubling of form elements is frequent at the stem level, where a distinction must be made with respect to grammaticalization. Doubling is frequently found in expressive forms, although it cannot be identified as a productive device of expressive word formation. What is decisive here is the filter of three radicals at the root level (cf. the grid in (12)) or, exceptionally, four radicals. Thus an etymological word formative consisting of two radicals would not pass this filter, unless doubled, thereby yielding a four-radical stem16. Most words with this formation have an expressive component, but not all, cf.17: (15) doubled root formation in Arabic °ws° : waswasa 'insinuate' °sr° : sarsar'cock' °dr° : dardar 'oak' : naTnaf 'mint' This formal characteristic cannot qualify as reduplication in the sense defined above, as the doubled formatives cannot be identified on the lexical and / or grammatical level. At the other extreme of the productivity scale of word formation is a device which affects the second radical. This is the most productive of the fifteen formation devices at the stem-W level (cf. 12) given by the grammarians of Classical Arabic (the so called second stem). Traditionally it is analyzed as (word-internal) partial reduplication18. With this stemformation, two meanings can be distinguished - albeit with some overlap in their use: (16) reduplication of second radical: a.

intensive formation (thus more or less expressive?): - °drb° 'hit': dar:aba 'to hit violently' - °ksr° 'cut': kas:ara 'to cut into small pieces' - °frq° 'separate': far:aqa 'to disperse'

b.

causative - °frh° - °hsn° -

formation: 'be happy': far:aha 'to make happy' 'be beautiful': has:ana 'to make beautiful' 'know': Bahama 'to teach'

Syntactic reduplication in Arabic

405

This traditional analysis presents a number of analytical problems. From a less abstract point of view, one which is not biased by Latinized transcription habits, the second radical in these forms is forticized. The Arabic term is ta/di.d "strengthening", a phenomenon which is not represented in orthography. The phonetic realization is a geminate with a clear differentiation between the implosive und the explosive parts, and thus not a sequence of identical consonants. Already McCarthy (e.g. 1982) analyzed this formation as augmentation by a consonantal mora on level 4a (stem-W) with the spreading of phonetic information from the second radical - in other words, not as reduplication. There is a clear difference between this formation and what are generally referred to as reduplicated roots: These are roots in which the second and third radicals are identical, such as °rdd° "to reply", and which are the result from trying to fit two-radical roots into the general three-radical template of Arabic roots. In the case of the "reduplicated roots", the phonetic realization depends on syllabification: the two identical radicals can be articulated both as geminates as well as separate consonants, whereas the geminate of the forticized second radical is never separated in the paradigm19. Thus in Old Arabic, only cases such as (10) and (11) qualify as reduplication in Arabic in the sense of this argument, not cases such as (13) and (14), and even less cases such as (15) and (16). In neo-Arabic varieties, the different types of doubling / reduplication are maintained and even elaborated20. As neo-Arabic varieties are used especially in oral language, doubling is common as expressive device, sometimes even lexicalized, cf. (17)

Neo-Arabic doubling a. Maltese 21 : al. xorob drink:PF.3.S.M a2. harg-u go.out:PF-3.P b. bl. b2. b3.

naqra little baxx low

Moroccan Arabic22: mdru.a mar:a saTa saia bhalbhal

naqra little baxx low

'he drank little by little' 'they went out discreetly'

'sometimes' (literaly: time time) 'from time to time' (literaly: hour hour) 'by the same way' (literaly: like like)

Doubling as a device of word formation is frequent, cf. the following Maltese examples, most of which are without an Old Arabic predecessor:

406 (18)

Utz Maas Doubled roots in Maltese capcap ferf er gerger harhar laqlaq

'clap' 'shake' 'growl' 'rattle' 'chat' etc.

Partial "reduplication" in the so-called second stem is especially productive with an innovative function, that of forming denominal verbs, a device used only marginally in Old Arabic, cf. in Maltese: xemx "sun" - xemmex "to sun", berqa "lightning" - berraq "to light", sabar "consolation" - sabbar "to console". In the following I will restrict myself to one especially productive case of syntactic reduplication in the domain of the verb phrase (cf. (9c)): Arabic has a special grammatical category, the masdar, which is used to articulate the complement of the verbal predicate. But before analyzing the construction, a clarification of the category masdar is necessary.

3.

Masdar formation

In most European grammars, verbal noun is given as the equivalent of masdar:; while older grammars refer to it an infinitive, which is rather confusing, as the masdar does not participate in the verbal paradigm (to be distinguished from the verbal system, cf. below). On the other hand, the masdar should be distinguished from deverbal morphological devices. In a certain sense it is a syntactically underspecified form, built directly from the radicals. It has the potential to share with verbs certain syntactic properties, especially valency: In Old Arabic it can govern its complements by the two adverbal23 cases: -u "nominative" for the main actant, and -a "accusative" for the secondary actant. However, it can also function as the head of a nominal phrase, governing an attribute marked by the adnominal case -i ("genitive"). Owing to its potential "verbal" properties, a masdar can be the head of a proposition but it cannot be the head of a narrative sentence ("verbal sentence"), as it cannot be marked for either ΤΑΜ nor for actancy (person). As the head of a secondary predication, a masdar is marked as the complement of the predicate - i.e. it is marked by case, just as other nominal forms are. Cf. (19a), with the masdar (?itia:m) as verbal object, and (19b) with the masdar darb as prepositional object24:

Syntactic reduplication in Arabic

407

(19) Masdar constructions (Classical Arabic)

a. Finna n-na:s-a PRES DEF-people-A

muham:ad-in

karih-u:

?Ma:m-a

condemn.PF-3.P

feeding-A

Tamr-an

xubz-an

masmu:m-an

Mohammed-G Amr-A bread-Α poisoned-A 'the people condemned Mohammed's giving Amr poisoned bread'

b. bi-darb-in

bi-s-sujuif-i

ru?u:s-a

qawm-in

by-hitting-G. [-DF] by-[+DF]-sword.P-G head.P-A people-G.[-DF] 'by beheading with swords the heads (accusative!) of people' So far the category is rather well defined, and has been maintained in the restructuring of the neo-Arabic varieties. As case affixes are no longer present, it is difficult to analyze the nominal forms, as there is no morphological difference between complements of a verbal form and attributes of a nominal form. Cases such as the following25 present the structure of regular noun phrases (by juxtaposition):

(20)

fjin

dl-xubz

knead.MD

[+DF]-bread easy

s Tib

sahal walakin at-tagrae but

l:i

[+DF]-flatten.M REL

fwija

difficult a. little 'to knead the bread (i.e. the dough) is easy, but to give (it) the flat form is a bit difficult' (20) is a case of the lexically restricted use of the /