Interaction of Derivational Morphology and Syntax in Japanese and English 9780429400933, 0429400934, 9780429684180, 0429684185

Table of contents :
Cover
Half Title
Title Page
Copyright Page
Original Title Page
Original Copyright page
Original Half Title
Acknowledgements
Table of Contents
Introduction
Chapter I. Some Theoretical Issues
1. The syntax/lexicon dichotomy --
The development of the Lexicalist Hypothesis
2. Word Formation Rules
2.1. Rule Formalisms
2.2. Level ordering and Adjacency Condition
2.3. Head
2.4. Inflection vs. Derivation
2.5. Semantics of Word
3. Typology of rules
3.1. Earlier attempts for rule typology
3.2. Rule vs. operation: Dowty's (1979) proposal
3.3. Interaction of processes. 3.4. Cliticization4. Lexicalism in Japanese Syntax
Chapter II. Deverbal Nominals and Compounds
1. English verbal compounds and argument structure
1.1. First Sister Principle
1.2. Verbal compounds vs. primary compounds
1.3. An alternative analysis of verbal compounds
1.4. No subject condition
1.5. The generic condition on the realization of the argument structure
2. Deverbal nominals and compounds in Japanese
2.1. Deverbal compounds
2.2. Suffixed deverbal nominals
3. A note on the condition on rendaku
Chapter III. Argument Structure and Derivational Morphology of Adjectives. 1. Argument structure of English adjectives1.1. Complement inheritance under nominalization
1.2. Adjectival compounds
2. Nominalization of Japanese adjectives
2.1. Two types of nominalizing suffixes
2.2. Argument structure and case marking of adjectives
2.3. Complements under nominalization
2.4. Extended domain of the nominalization
3. On the verbalizing suffix-garu
Chapter IV. Phrasal Suffixes I: Alternating Case Marking
1. Phrasal suffixation and reanalysis
2. Facts about Japanese stative predicate constructions
3. Proposal
4. Morphological transparency. 5. Conditions for the reanalysis5.1. Distance between the second NP and the predicate
5.2. Conjunction and comparatives
5.3. The distance between the verb and-tai
5.4. Semantic conditions
6. Concluding remarks and some issues
6.1. Basic and derived case patterns for stative predicates
6.2. On the object-hood of the second NP
6.3. The accusative NP condition --
a problem
Chapter V. Phrasal Suffixes II
1. Suffixes with phrasal scopes
1.1. Reciprocal suffix -au
1.2. On the suffix -sugiru
2. Nominal forming suffixes
2.1. Adjectival nominal forming suffixes. 2.2. Predicate nominal forming suffixes3. Subordinate clause suffixes
4. The domain of passive suffixation
Chapter VI. Concluding Remarks
Bibliography.

Citation preview

ROUTLEDGE LIBRARY EDITIONS: JAPANESE LINGUISTICS

Volume 3

INTERACTION OF DERIVATIONAL MORPHOLOGY AND SYNTAX IN JAPANESE AND ENGLISH

INTERACTION OF DERIVATIONAL MORPHOLOGY AND SYNTAX IN JAPANESE AND ENGLISH

YOKO SUGIOKA

First published in 1986 by Garland Publishing, Inc. This edition first published in 2019 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 52 Vanderbilt Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 1985 Yoko Sugioka All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: ISBN: ISBN: ISBN:

978-1-138-36949-8 978-0-429-40043-8 978-0-367-00174-2 978-0-429-40093-3

(Set) (Set) (ebk) (Volume 3) (hbk) (Volume 3) (ebk)

Publisher’s Note The publisher has gone to great lengths to ensure the quality of this reprint but points out that some imperfections in the original copies may be apparent. Disclaimer The publisher has made every effort to trace copyright holders and would welcome correspondence from those they have been unable to trace.

Interaction of Derivational Morphology and Syntax in japanese and English

Yoko Sugioka

Garland Publishing, Inc. • New York & London

1986

Library of Congress Cataloging-in-Publication Data Sugioka, Yoko. Interaction of derivational morphology and syntax in Japanese and English. (Outstanding dissertations in linguistics) Originally presented as the author's thesis (Ph.D.)-University of Chicago, 1984. Bibliography: p. 1. Japanese language-Morphology. 2. Japanese language-Syntax. 3. English language-Morphology. 4. English language-Syntax. 5. lexicology. I. Title. II. Series. Pl559.S8 1986 425 85-31112 ISBN 0-8240-5478-4 (alk. paper)

© 1985 by Yoko Sugioka All Rights Reserved

The volumes in this series are printed on acid-free, 250-year-life paper. Printed in the United States of America

THE UNIVERSITY OF CHICAGO

INTERACTION OF DERIVATIONAL MORPHOLOGY AND SYNTAX IN JAPANESE AND ENGLISH

A DISSERTATION SUBMITTED TO THE FACULTY OF THE DIVISION OF HUMANITIES IN CANDIDACY FOR THE DEGREE OF DOCTOR OF PHILOSOPHY DEPARTMENT OF LINGUISTICS

BY YOKO SUGIOKA

CHICAGO, ILLINOIS DECEMBER, 1984

ACKNOWLEDGEMENTS I

wish first

to express

members of my committee, Bill Darden, the

for

dissertation

graduate

my profound

Jim

McCawley,

research

benefited enormously

and

the

University

from the

McCawley.

His

persistent

various stages of

of Chicago

and

insightful

linguistics taught

Jim

During

was consistently

I am truly thankful for all this.

I am

especially grateful to

help

he gave

the reading

have

from Jim

is to be a linguist.

my dissertation writing,

shaping

theoretical framework

I

my

on various

sessions I took

tackling many problems of

encouraging and helpful.

Throughout

enthusiasm

me a great deal about what it

out of

and

numerous courses

a number of reading

in

Jerry Sadock,

writing.

subjects and

the time of

the

guiding me through the

education at

approaches in

gratitude to

Jerry Sadock for

the enormous

this dissertation.

of this dissertation sessions I

had with

The

basic

developed first Jerry.

He

also

influenced me a great deal through his past and ongoing work on Greenlandic cannot

and the

thank him

consistent

morphology/syntax interaction.

enough for

encouragements

versions of each chapter. for the insightful

the

he

detailed criticisms

offered me

on

the

I and

earlier

I am also grateful to Bill Darden

comments on the earlier

i

version of this

work,

as well as for the help he gave me as the chairman of

the department. I would like to thank the at

the

Department

faculty and my fellow students

of Linguistics

Chicago for

providing such a

study in.

I owe

friend, in

of

warm and cozy

big thanks to

University

of

environment to

Rachel Lehr,

my special

who discussed with me many of the issues brought up

this dissertation

and

provided me

friendship whenever I needed it.

as for lots of encouragements. Farkas,

Kat Momoi,

support.

I

I

great

also wish to thank Donka

on various

Finally,

such a

the manuscripts as well

John Richardson,

stimulating discussion

with

I am also grateful to Ann

Wehmeyer for reading and correcting

moral

the

and Rob Chametzky for

linguistic subjects

would

like

to

thank

and my

non-linguist friends in Hyde Park for making my life here so much more enjoyable and meaningful. organizations provided me

The following institutions and with financial support during my

graduate study at Chicago:

the Division of Humanities of the University of Chicago, the Center for Far Eastern Studies of the University of Chicago, Rotary

International,

Association Association.

of

P.E.O.

University

Women,

International, Kobe

College

American Alumni

I am deeply grateful for their financial help,

without which I could not have completed my graduate work at Chicago.

ii

Finally, parents, Todokoro,

I wish

to

express my

Tomio and Setsuko Sugioka, for

their consistent

material.

iii

deep

gratitude to

and my husband,

support,

both

my

Kazuo

mental and

TABLE OF CONTENTS Page ACKNOWLEDGEMENTS INTRODUCTION

1

Chapter I. SOME THEORETICAL ISSUES . . . . • . • • 1. The syntax/lexicon dichotomy--The development of the Lexicalist Hypothesis • • . • • • . • • • 2. Word Formation Rules • . . . • • 2.1. Rule Formalisms • • • . . • • 2.2. Level ordering and Adjacency Condition . . . . . . • 2. 3. Head • . . . . . • . • • 2.4. Inflection vs. Derivation 2.5. Semantics of Word • . • . 3. Typology of rules • • • • • . 3.1. Earlier attempts for rule typology 3.2. Rule vs. operation: Dowty's (1979) proposal • . . . . . . . • 3.3. Interaction of processes 3.4. Cliticization • • • • • . 4. Lexicalism in Japanese Syntax II. DEVERBAL NOMINALS AND COMPOUNDS • . 1. English verbal compounds and argument structure . . . . • • . • . . 1.1. First Sister Principle • • • . • 1.2. Verbal compounds vs. primary compounds . • • • • . • • . • . 1.3. An alternative analysis of verbal compounds . • . . . . . . • 1.4. No subject condition . • . . 1.5. The generic condition on the realization of the argument structure • • • • . 2. Deverbal nominals and compounds 1n Japanese . . . . • • • . • . • • 2.l.Deverbal compounds • . • • . • . 2.2. Suffixed deverbal nominals 3. A note on the condition on rendaku

v

4

4 8 9 • 13 17 19 21 26 26 27 30 42 47 56 56 56 58 64 72 73 77 79 93 105

III. ARGUMENT STRUCTURE AND DERIVATIONAL MORPHOLOGY OF ADJECTIVES • • • • • • • • • • • • • • • 1. Argument structure of English adjectives 1.1. Complement inheritance under

nominalization • • . . . • . . . • ••••••• 2. Nominalization of Japanese adjectives • 2.1. Two types of nominalizing suffixes 2.2. Argument structure and case marking of adjectives • • • • • • • • • • 2.3. Complements under nominalization •• 2.4. Extended domain of the nominalization 3. On the verbalizing suffix -garu • • • • 1.2. Adjectival compounds

IV. PHRASAL SUFFIXES I: ALTERNATING CASE MARKING 1. Phrasal suffixation and reanalysis

2. Facts about Japanese stative predicate constructions • • • • • • 3. Proposal • • • • • • • • • • • • • • • 4. Morphological transparency 5. Conditions for the reanalysis • • • • • 5.1. Distance between the second NP and the predicate • • • . • • 5.2. Conjunction and cpmparatives 5.3. The distance between the verb and -tai • • • • • • • • • • • • • 5.4. Semantic conditions • • • • • • • • 6. Concluding remarks and some issues • • • 6.1. Basic and derived case patterns for stative predicates • • • • • • 6.2. On the object-hood of the second NP 6.3. The accusative NP condition--a problem • • • • •

V. PHRASAL SUFFIXES II • • • •

114 115 116 123 126 126 130 132 137 146 153 153 155 158 161 165 165 167 172 173 176 176 181 184 190

1. Suffixes with phrasal scopes 1.1. Reciprocal suffix -au •• 1.2. On the suffix -sugiru • . 2. Nominal forming suffixes ••.•• 2.1. Adjectival nominal forming suffixes •

191 191 199 209 209

2.2. Predicate nominal forming suffixes 3. Subordinate clause suffixes • • • • • • • • 4. The domain of passive suffixation ••••

211

VI. CONCLUDING REMARKS

214 220 229

BIBLIOGRAPHY

237

vi

INTRODUCTION issue that is addressed

The main

question of how the proper syntax should be

defined.

formation processes involve

some

boundary between the lexicon and In this respect,

in Japanese and

interaction

is the

in this thesis

of

various word

English which

morphology

and

seem to

syntax

are

examined here. In the past

15 years of studies on the

formation

we saw

starting

with

a

development

Chomsky's

lexicon and word

of so-called

'Remarks

on

Lexicalism,

Nominalization',

followed by various works by people such as Aronoff, Siegel, Allen, Lieber, Selkirk, Kiparsky and so on. it is

maintained that

lexicon and does not rules.

In

other

word formation

In these works

belongs only

systematically interact with syntactic words,

morphology,

sub-word-level derivations, is never

which

fed by

relatively few problems for

involves

syntax,

involves phrase and sentence level derivations. this view creates

to the

which

Even though English word

formation, that is not the case with typologically different languages.

For

Greenlandic,

a polysynthetic language,

be

syntactic

by

instance,

Sadock

noun

(1980}.

1

incorporation

in

West

has been claimed to As

an

agglutinative

2 language, Japanese should serve as a good testing ground for the

universality only a

however, further

of

more,

the Lexicalism.

few studies

the

arguments against

lexicalist them)

such constructions

There

done in claims

have pretty

have

this respect,

and

about Japanese

(or

much been

limited to

as passives and causatives,

been well discussed

been,

in syntactic analyses.

which have

Thus

1

one major

purpose of this

dissertation is to examine a

wide range of

word

processes

question

formation

plausibility of

in

the lexicalist

Japanese

and

hypothesis as

a theory

the of

universal grammar. The

organization

following.

formation processes

as a framework of

In Chapter II we look at

differences.

Japanese,

is

as

the

development of the lexicalist

rule typology approach to

dichotomy is proposed

English and

dissertation

the various specific claims

regard to word

after which a

this

In Chapter I the

hypothesis and made in

of

and assumptions are discussed,

the syntax/lexicon this dissertation.

deverbal nominals and compounds in and

discuss their

similarities and

Chapter III will deal with nominalization and

verbalization of adjectives in the two languages and discuss in particular the important role argument structure plays in morphological derivations. consequence

of the

Chapter IV takes up one specific

framework

proposed

here,

namely

the

1 0ne notable exception is Kageyama (1982), and a number of claims he makes are discussed in many places of this dissertation.

3 possibility of analysis of Japanese.

and apply

the derived stative predicate In Chapter V,

are discussed, affixes with

phrasal suffixation,

it to

the

constructions in

other suffixes with phrasal scope

and some of them are compared to the English similar functions.

The observations

these chapters are summarized in Chapter VI.

made in

CHAPTER I SOME THEORETICAL ISSUES This

chapter

theoretical

provides

issues in

lexicon in grammar general.

In

the

the

basic

background

treatment of

and the properties of

Section

1

will

we

word formation in discuss

the

the boundary between

In Section 2

the lexicon.

the

role of · the

briefly

development of 'lexicalism' concerning syntax and

the

for

we will

discuss a

number of claims that have been made about the properties of word formation and the organization of the lexicon; them are

specifically for

English,

supposed to hold across languages. claims will

be discussed.

proposed that the

but

Then in Section

notion of 'rule typology'

of the

them are

Some problems for those

syntax/lexicon dichotomy is one fruitful realistic view

many of

some of

interaction of

3 it

will be

applied to the

way of acquiring a the two

components.

Finally, Section 4 will have a brief survey of recent trends in the study of word formation in Japanese. 1. The syntax/lexicon dichotomy--The development of the Lexicalist Hypothesis The basic unit is a word in the lexicon and a sentence in syntax.

Words are

whereas sentences

usually memorized

are usually generated

4

and

used as

each time

such, for the

5

utterance.

Noting

'Prolegomena

to

this,

Word

Halle

Formation'

writes

(1973):

fundamental difference between the use

in

'There

his is

a

of words and the use

of sentences; one encounters new sentences but not new words all the time'. lexicon

is

Thus it is a well-established view that the basically

a

list

of

words

with

their

idiosyncratic properties while syntax is a set of productive rules that combine them and generate sentences. contains memorized

knowledge while

The lexicon

knowledge of

syntax is

productively used. If it were

the case that in natural

not analyzable in any regular way, lexicon

and syntax

grammar.

would be

In reality,

from that simple. free and bound

this simple view of the

adequate for

however,

the

that aspect

of

picture we get is far

Words are often made up of smaller units,

morphemes,

form complex words. character

languages words are

and words themselves

combine to

The processes of word formation vary in

from quite

irregular and

idiosyncratic ones

to

mostly regular and productive ones.

A problem then arises

as

processes

to

whether

such word

formation

should

be

relegated to the lexicon or syntax.

The basic unit, however

complex it

although

somewhat

is,

is

resemble

productively

used.

although there is

still a those To

word, of

take

syntax an

in

example

no doubt that washable is

the processes

that from

they

are

English,

a single word,

it is part of an English speaker's knowledge that the suffix

6 -able attaches meaning

to a

'can

be

verb and forms

V-ed'.

productively used to

This

an adjective piece

of

with the

knowledge

form or understand a word

is

of the form

v-able. As

long

as

components,

we

the

maintain that

there

lexicon and syntax,

exist

these

two

which seems to

be a

reasonable view, there are two basic approaches to take: one can either assign word formation component,

even though

idiosyncracy which lexicon,

or

they

processes to the syntactic exhibit

still have to

alternatively,

work of

relevance to this

literature

is

Robert

assume that

the lexicon.

issue in the

B.

Nominalization (1960),

be somehow listed

one can

formation processes are part of

various degrees

Lees'

The

of

in the all word

The earliest

generative grammar Grammar

of

English

which took the former approach.

At

that time any kind of regular relationship between two forms was

to be

expressed by

approach,

often

syntactic transformations.

referred

to

as

This

'transformationalist

hypothesis' has the advantage of keeping the lexicon only to idiosyncratic

information while

regularities by syntactic rules. on nominalization' (1972)

such

as

Chomsky,

perceivable

in his 'Remarks

questioned this approach.

on some differences between nominals

capturing all

the

derived nominals vs. morphological

and

Based

gerundive semantic

idiosyncracies and the surface NP structure of the former as opposed to

the latter,

he

proposed that

derived nominals

7

should be derived proposal, the

in the lexicon.

which is called

idea that

systems,

and

relegated

grammar

The

the 'lexicalist hypothesis',

consists of

a

that the regularities

solely

to

the

set of

transformations

'a demarcation problem' in the 1972 idea

figures

but

not be

should

be

He called this task

article~

more prominently

is

interacting

we find should

distributed among different components.

this

insight behind his

in

more recently,

his

work

(e.g.,

Chomsky 1981) under the name 'modularity'. It is important to note here that Chomsky (1972) proposed the

'lexicalist

nominals

but

hypothesis' not

all

to

deal

derivational

with

the

derived

morphology.

His

proposals, however, have been extended and strengthened by a number

of linguists

in

Jackendoff

(1972,

morphology

should be

the

1976)

development of claims

placed

in

that

'lexicalism'.

all

derivational

the lexicon,

where

any

morphological or semantic regularities between lexical items are

captured by

redundancy rules.

Aronoff (1976)

also

claimed that all derivational morphology is in the domain of the lexicon.

Recently,

even inflection is claimed

performed in the lexicon.

(See Lieber 1980,

Lapointe

1983,

People subscribing

maintain

the

etc.).

'lexical integrity

to be

Kiparsky 1982, to this

principle',

whereby

view no

morphological operation is allowed in

the domain of syntax.

Thus Chomsky's

has yielded

years

some

Lexicalist Hypothesis

quite

strong claims

between syntax and the lexicon.

regarding

the

over the dichotomy

8 2. Word Formation Rules Once the word-level

derivation was considered not

part of syntax but of the lexicon,

to be

it was a natural move to

capture the regularities of morphological derivation by some form

of rules

Jackendoff

which

(1976)

are

separate from

proposes

syntactic

morphological

and

rules. semantic

redundancy rules,

and he maintains that creation is not the

primary

of

the

lexicon,

rules

can

be

role

redundancy

understanding new

words.

although

used It

in

he

creating

was Halle's

says as

well

that as

'Prolegomena to

Word Formation' (1973)

that "rediscovered" and brought back

to

morphology

light derivational

generative

grammar.

formation was

1

His

as

program

for

in

word

Aronoff

(1976), Siegel (1974), Allen (1978), Lieber (1980),

Selkirk

and

beginning

so on. to be

structure as

number of

studies

in

linguists;

(1982)

followed by a

productive rules

Through

considered

their work word to

a subcomponent of

section discuss several basic

have much grammar.

formation is

complexity We will

Under each issue alternative views some problems

specific

claims

discussed in later

with them concerning chapters,

grammar.

will be briefly surveyed

will be

pointed out.

specific

processes

after which we

how the data examined bear with those claims.

1

in this

issues concerning the overall

properties of word formation rules and its role in

and

and

See Lipka (1975) for relevant remarks.

More will

be

will evaluate

9

2.1. Rule Formalisms There have been basically three what

form

word

formation

types of proposals as to should

transformation, phrase structure rule,

take;

lexical

and no rule but only

lexical insertion.

?.!.1. Lexical Transformation In

order to

compounds in

capture certain

English,

Roeper

and Siegel

lexical transformation rules. of introducing (1973).

generalizations on (1978)

verbal proposed

They attribute the basic idea

transformations in

In their analysis a

the lexicon

to Vergnaud

compound like coffee maker is

derived by a transformation in the lexicon as follows. (1) [make]V [coffee]N ---> [coffee maker]N The first

sister to

attached to affixed.

the verb

the left of

in its

the verb at

the same time

-er is

In a different paper(Keyser and Roeper 1983) it is

claimed that this rule is an in the lexicon.

instance of the rule

shown by

'move-~'

It is also pointed out there that this rule

operates on structural terms rather This is

subcategorization is

the fact that

than on thematic roles.

nouns of

various thematic

roles can appear as the first element of the compound:

2

(2) home-cooked (loc.), slave-built (agent), razor-sharpened (instr.)

2 See Levi (1976) for detailed observations on the semantics of such compounds as well as a large collection of data.

10

What the first element of these

compounds share is the fact

that they would appear in a phrase following the verb (if we ignore the referred to

as 'the first

Siegel, is structurally (1977),

sister principle' by

defined,

the rules whose domain

thematically

defined

which is

This generalization,

prepositions).

are

and

Roeper and

according to

Wasow

is structurally rather than

transformational

rather

than

lexical.

Note that Roeper and Siegel refer to the rule as a

'lexical

transformation'

transformational rule

in

the

sense

operating in

that

this

the lexicon.

is

a

We will

discuss the apparently paradoxical status of this rule as so defined in the

sections below.

compounds will be

More details

taken up in Chapter II.

note here that transformation is rule that because

has been proposed of the

It suffices to

one form of word formation

for deriving

structural

about verbal

nature'of

verbal compounds

the

domain of

this

putative rule. 2.1.2. Phrase Structure Rules Jackendoff (1976) notes that "the syntactic analogue of a morphological

redundancy

transformation." pursued

by Selkirk

This

rule idea

(1982).

is

has She

a been

PSR most

defines WFRs

and

not

a

extensively as the

X'

version of context-free PSR, which she calls 'Word Structure Rules'. with

WSRs define the structure of a word in a tree form

the initial

symbol

XJ(word),

which may

expand

to

11

Jf.

x1 (stem),

symbols

intended to be

(root),

and

affixes.

WSRs

generative devices for new words

are

as well as

redundancy rules for existing words. The

idea

that

formation)

WFRs

rules are

and

syntactic

formally

the

(i.e.,

sentence

same except

for

the

difference in the levels of X' is an attractive one, but the question

how far

empirically.

the

from

has

to be

tested

For instance, Selkirk's WSRs are based on the

assumption that all far

parallelism goes

clear

word structures are headed, in

sentence

structure

but it is

that

all

the

constituents are headed.

It is also important to note that

she claims (Selkirk 1982:

3)

sufficient to only part of languages.

that although these WSRs are

characterize English

word structure,

a typology of word structure We can immediately think

it is

systems of human

of a few processes of

morphology which cannot be straightforwardly expressed with WSRs;

namely,~-derivations,

dvandva constructions.

reduplications,

and

Another problem with this system is

the status of the category given a category

umlaut,

status,

'affix'.

Although affixes are

this category does

not belong to

the X'-hierarchy with other categories according to Selkirk. Since affix is a major element affixes are known Carlson and status

to function semantically as

Roeper 1979),

undermines

in word formation,

the

this lack

adaptation of

and some

adverbs (see

of account

of their

X'

and

system

the

putative parallelism between syntactic and word structure.

12 2 .1. 3. No WFRs Lieber (1980)

claims

that there is no need

affixation and compounding.

Instead, she postulates lexical

structures

of an

items

inserted according

are

frame.

unlabeled tree

node)

which features

of the head

node. gives

to

their

following

affixation;

(b)

the category of

percolation through

are copied onto

points

affixation

as

advantages

(a)

nation-al-iz-ation,

never

of

who

lexical

It unifies compounding

Percolation

is

the dominating

by Kiparsky (1983),

is

needed

features triggering phonological processes; why

lexical

subcategorization

(i.e.,

This approach is supported the

into which

is determined by

insertion approach over WFRs: and

form

The hierarchical structure

the dominating

for WFRs of

for

(c) It explains

extrinsically

organ-iz-ation-al)

anyway

ordered

(d)

It

(see is

in

accordance with the fact that features that trigger WFRs are the features of the stem. We

have

seen

that

formalism

for word

important

questions

the

problem

formation

is

that remain

there are between syntactic and

also important to

any rule formalism

note

appropriate

quite open-ended. are:

what

The

what

in word formation rules.

here that there

that can be claimed- to

of word formation across languages.

rule

parallelisms

word formation rules;

regularities should be represented It is

of

hasnrt been

handle all types

13 2.2. Level ordering and Adjacency Condition One prominent claim

of the recent studies

is the so-called level-ordering assumes that

of morphology

hypothesis.

morphological processes are

This approach

ordered according

to the levels they belong to. It

was first

proposed by

affixes (class 1)

Siegel

(1974)

that

certain

which affect the phonological form of the

base morpheme and do not

tend to have transparent semantics

never occur outside those without such properties (class 2). This

idea was

adopted and

extended by

Allen,

compound formation as class 3 processes.

who

added

Allen (1978) also

proposed the Adjacency Condition, under which WFR can affect only items in the adjacent morphological cycle, WFR

'blind'

to

the

internal

structure

or

thus making derivational

history of the lexical items. This

development

Kiparsky (1982,

was

1983),

Morphology'.

Kiparsky

recently who

taken

calls his

assumes

rules.

three

distinctive

that

idiosyncratic and

are are

formally ordered

transparent,

and

for

level

and a

Condition governs the application of

of

and he

English.

One

is the postulation that

semantically

before

by

a set of phonological

levels

noteworthy insight in his framework

general

each

The output of each level is a lexical item,

postulates

rules

further

framework 'Lexical

that

morphological rules is accompanied by

up

rules principle

specific that of

these rules.

are

and more

Elsewhere He also

14 adopts Allen's Adjacency Condition by using brackets to mark the output of each level and postulating the Bracket Erasure Convention,

which erases

end of the level, on

a

all the internal brackets

thus making

certain level

to

at the

it impossible for the rules

have

access to

the

derivational

history of a lexical item. Although

his framework

makes

which hold true for the English some problems.

a

number of

data,

predictions

there seem to remain

Most notably, there exist at least two types

of violation of the level ordering he assumes, and he has to either

allow some

limited

bracket erasure for the

level

recursion

such cases.

ordering

is

of levels

One type

exemplified

by

or

block

of violation of a

word

like

analyzability, in which a level 2 affix able occurs inside a level 1

affix ity,

and

a word like

which a level 2 affix un has level

1

affix

ity

subcategorization of

ungrammaticality,

in

to attach to the stem before a

does,

in

the affix

order

to

un(/__Adj.).

meet These

the cases

have led some people to separate morphological structure and semantic structure. (See Williams 1981, Lieber 1980, Selkirk 1982.)

Further in this direction, Pesetsky (1983)

a mapping

operation between

Form which takes Kiparsky

(1983),

lexical structure

the form of the on

the other

proposes

and Logical

'move-c{' transformation. hand,

proposes that

the

bracketing paradox of a word like ungrammaticality should be accounted for as an iristance of morphological reanalysis.

15 The

second

type

of

level violation

affixation to a compound or a phrase.

has

to

do

with

It is exemplified by

such words as transformational grammarian, matter-of-factly, shirt-sleeved,

and

so on.

In view

of such

examples as

these, Kiparsky says that a limited recursion from syntax to the lexicon must be allowed. how 'limited' these the productive

One has to wonder,

exceptions could be when

formation of adjectives

however,

one considers

like shirt-sleeved,

kimono-sleeved, long-sleeved, French-sleeved, and so on, take one example number

of such

from English. instances of

Furthermore,

productive processes

English found with polysynthetic languages.

This

point

is

there

to

are a outside

languages or agglutinative crucial

in

level-ordering approach to word formation.

evaluating

the

This will be a

major topic of discussion in the following chapters. Partly above,

motivated by Aronoff and

such

counterevidence as

Sridhar (1983)

recently proposed

abandon the level- ordering hypothesis the

classification 'word

assuming any defined

facts can be

ordering between

as attaching

'stem affix' claimed

affix' and

to major

discussed

and to adopt instead 'stem affix'

the two.

without

'Word affix'

lexical categories,

attaches to morphemes.

to

They claim

is

while that the

to follow from the level-ordering hypothesis

accounted for by

two types of affixes and of Elsewhere Condition.

the different properties

of these

the independently needed principle They also claim that the bracketing

16

paradox assume

problems with different

structure

morphological word, above

the

level

level ordering for

disappear once

phonological

which they point of

word

in

the

they

word

and

out has been accepted assignment

of

rhythm

structure. In spite of the problems,

there seems to be an intuitive

idea behind the level-ordered

morphology,

articulated by Kiparsky's framework;

namely, in the process

of building words, phrases, and sentences, processes are

more regular in form

closer to the sentence level. have syntactic processes.

On

which is clearly

we find that the

and meaning,

as

it is

this end of the scale we

Going in the opposite direction,

we find the processes to become more and more irregular both in form and meaning, and finally, the forms (morphemes) further unanalyzable and simply have have the base that,

in

dealing with word

somewhere

in between,

regularity. have

of the lexicon.

we

It

out in

the

perceive

which lie

different

levels

it is also true,

past,

here we

therefore seems natural

formation processes

At the same time,

pointed

to be listed;

are

that there

of

as people

exists

some

evidence against the strictness of

such ordering of levels.

We have

at this moment,

should

to leave this note

one

implications do hypothesis and

issue open

important

question.

the predictions made by the problems with it

syntax/lexicon dichotomy?

That

is,

but we what

the level-ordering

have for the

issue of

This very question seems to have

17 been

neglected because

been

discussed

position.

with

We

assumption in

the the

should,

level-ordering hypothesis assumption

however,

of

the

lexicalist

leave

the

lexicalist

order to properly investigate

particularly in

view of the

has

this question,

following facts.

The higher

level word formation processes

have very similar properties

to

formation

phrase (i.e.,

regularity and

syntactic)

phrases.

Further,

compositionality.

bracketing paradox extend

in

terms of some

above the word level

the

cases of

and involve

And finally, if we are to recognize a phonological

word to have different structure

from a morphological word,

as Aronoff and Sridhar (1983) suggest, the phonological word in their definition would include genuine cases of a phrase, even of non-constituents. 2.3. Head 'Head' is an important notion The majority of

in syntax. can

be

considered

definition,

of

words and phrases contain what

head.

xn

the head of

According

frameworks

to

the

shares the features with

a sub-word level to

varying from number

as

both in word formation and

with

basic

xn,

In a

a phrasal level.

a version

of

the

X'

system

(various works in GPSG, Lieber 1981, Selkirk 1982, etc.), mechanism called 'feature percolation' the feature

of a particular

In word structure,

each

item

lexical

must

be

a

is employed by which

element is transferred

mother node.

X

to the

the relevant features of

properly

percolated

to

the

18 dominating node, but it

and become the features of the whole word,

has not been completely

agreed upon as to

how that

can be always carried out. English depending

compounds on

can

whether

classified

be

are

they

into

headed

or

two

types With

not.

endocentric compounds in English, it is always the case that the

rightmost

element

is

the

head.

With

exocentric

compounds,

such as [pick pocket]N,

neither of the items is

the

There

of non-headed

compound

called 'dvandva compounds',

which consist of the

items of

the same

where both

head.

is another

lexical category,

type

items can be thought of as

head.

not very common in English,

but

discussion on

Japanese in

of the

compounded

This type of compound is in some languages (see the

Kageyama 1982)

it is

a fairly

productive process. As for affixation processes, all the affixed items can be considered as simple

headed,

but

designating the

It

has

been

matter.

assumed

category-changing affixes are the head most suffixes are

head is by

many

of the word.

category-changing while only a

not a that Since

couple of

prefixes are (en-, be-) in English, there is a tendency that the rightmost element with

compounds.

is the head; A

problem

affixation comes into the carried by

arises

picture.

inflection such as number

percolated to the mother node to

which

is also observed when

inflectional

Clearly,

the features

and tense have

to be

become the features of the

19 word

itself,

and

yet the

inflectional

determine the

lexical category of

fact

to

has led

(1981)

has treated them as

opposed to that

issue,

inflectional

in

the

do

word.

the

Williams

while Selkirk (1982)

This problem actually which we

affixes

not This

treatment of

of their headhood. head,

approach.

another important whether

the entire

a discrepancy

inflectional affixes in terms

affixes

turn to

should

be

is

bears on

next,

namely,

treated

in

the

lexicon along with derivational affixes. 2.4. Inflection vs. Derivation So far

there seems to

be no solid

for distinguishing inflectional processes.

criteria established

processes from derivational

(See Anderson 1982.)

Nevertheless,

assumed that they are distinguishable

it has been

for a majority of the

cases, the difference being that inflection forms a paradigm of one lexical item,

while derivation derives a new lexical

item

one.

from

derivation

an

old

used to

places of grammar. autonomy

of word

Furthermore,

be considered

to

inflection

belong in

and

different

Even within the lexicalist approach the formation

from

syntax was

claimed

for

derivational morphology and not for inflectional morphology. (See Chomsky 1972,

Aronoff 1976.)

More recently,

it was

proposed that inflectional morphology be treated in the same way as derivational morphology,

extending the domain which

the lexicalist hypothesis covers, while some people maintain inflection and derivation should be dealt with separately.

20 It

has

been

noted

by Lieber

morphological processes,

(1979)

umlaut for instance,

both in inflection and derivation, processes use German

that

for

conversion sometimes

participle stem of the verb such as Fund V,N. has led

Lieber to include inflection This view is shared

lexicon.

(1982), and Lapointe (1983). hand,

takes inflection

to be

outside

past

This finding of the Selkirk

the lexicon.

"whether

or

uses the

in

Anderson (1982), on the other

further

inflectional

example,

by Williams (1981),

be "what is relevant

that

often figure

in the domain

defines inflection to says

same

and certain derivational

the inflectional paradigm;

a verb-to-noun

the

derivational

a

certain

depends

on

He

to syntax." category how

much

He is the

category realized in inflection is integrated into syntactic principles," and notes that what counts as inflectional is a language specific

matter.

We

should note

here that

the

validity of Anderson's program depends on the clarity of the notion 'relevance to syntax'. The

controversy

stems

somewhat dual character of words

in

terms

of

at

inflection;

morphology

interpretation and distribution syntactic matter.

least

and

partially

from

the

the process affects phonology,

but

of the inflected word

the is a

For instance, the nominative case may be

realized by affixation,

while lexical insertion of

the NP

into the subject position of a sentence must be performed on a syntactic

tree.

We

will discuss

a new

approach below

21 using the notion of rule typology, with which we will try to give an alternative to these existing analyses. 2.5. Semantics of Word Since some words are naturally arises as them.

structurally complex,

to how form and meaning

More specifically,

the question

regularities of word formation can hierarchical

structures,

is

the question are related in

is,

if

certain

be captured by assigning the

semantic

structure

isomorphic to the word structure? This

question

became

ordering hypothesis. condition this

more prominent

(See 2.2.)

approach does

under

the

level

Coupled with the adjacency

not allow

a derived

word to

refer to the structure it had on a previous level, resulting in a strong

prediction as to the way

and can have

semantic scope over each

semantic structure levels.

Allen

is isomorphic with (1978),

in

lexical items combine other,

provided the

the one

pursuing

this

assigned by restrictive

approach, runs into a number of problems with the semantics. The

most serious

structure Since in

cases arise

involves her scheme

when

prefixing or

the putative

suffixing

compounding belongs

to

semantic compounds.

to level

3 while

affixation belongs to level 1 or 2, compounds cannot undergo affixation as example,

a whole,

hence some

problems arise.

For

the meaning of a verbal compound like story teller

depends crucially on the information that the second item is

22 derived

from a

verb

tell

object such as story,

which subcategorizes

a

direct

but when the compounding takes place,

([story]N + [teller]N), that information cannot be available under

the adjacency

condition,

tell+-er belongs to level 2. that the

since

Allen

the affixation

was then forced to say

semantic interpretation rule for

(basically

the same

Principle used

as Roeper

as an

level ordering.

and

of

verbal compounds

Siegel's First

interpretive device)

Sister

does not

obey

Put more plainly, she abandons the idea of

having semantic structure isomorphic to lexical structure in word formation. On the

other hand,

Allen

approaches a

similar problem

with an adjective forming suffix -ed quite differently. possible, and semantically straightforward,

One

account of this

suffix involves -ed attaching to X, X being aN (i.e., noun, compound noun, adjective+ noun), yielding an adjective with the meaning 'having X, or bearing the characteristics of X', as exemplified by the following: (3) [pencil point]-ed, [two leg(g)]-ed, [cold blood]-ed. Such

an account,

however,

adjacency condition,

be

allowed under

which Allen maintains.

claims that there is no to be even

cannot

the

She further

justification for requiring X above

a semantic unit,

not to mention

a lexical one,

since X is not always a compound but can be a phrase as well (see two phrase

legs,

cold

blood above),

is "theoretically

improbable"

and suffixation according to

to a Allen.

23

The significance Thus

for these

convincing

of this complex

semantic

insufficient

one,

remark will adjectives

account

namely,

semantic structure

she cannot

but that

[cold]A +

be discussed

can 'cold

later.

provide

only

a

give

an

blooded' has

the

[blooded]A,

with

the second

adjective modified by the first one, whatever that means. Williams (1981) between possible

also points

out several

lexical structure and

discrepancies

semantic structure,

as exemplified by the following: (4)

[hydro electric]-ity, [Godel-number]-ing, [atomic scient]-ist

Williams

tries to

notion of

give

some account

'lexically related'

by

whereby Y

proposing a in the

new

following

tree can be related to X: (See Williams 1981 for details.)

X

4-'-1 This

notion

explicitly provision

of

lexically-relatedness

notes,

clearly

he maintains

non-compositional.

Allen's

cannot precede any affixation. with exist.

this new The

notion

of

as

is,

position that It

With

this

compounding

is noteworthy that even

relatedness some

example Williams

Williams

gives is

exceptions

do

'reair-condition',

with which there is no way one can relate 'air-condition' to the whole word;

for this case he is forced

'air-condition' has been reanalyzed as a stem.

to assume that

24

Selkirk

(1982)

agrees

with Williams'

structure is non-compositional,

although

claim that

she contends that

compounding can precede level 2 affixation. to treat

a case

like set theoretic

[set]#[theoret+ic],

with -ic

leads her to claim that when compound

the structure

word

This allows her

to have

the structure

as level 1 suffix.

It also

level 2 suffixes appear with a

can be

ambiguous,

as

illustrated

below: (5) [bath room]-less I

[bath] [room-less]

[head strong]-ness I It is

[head] [strong-ness]

claimed that both structures

interpretation,

following

the

can be 'given principle

the same

allowing

for

non-compositionality in the interpretation of word structure suggested by Williams'(Selkirk 1982: seems

absurd,

for

structures. some suffixation,

for unnecessary

alternative

Once we

allow compounding

to precede

a word like 'bathroomless' can be given a both derivationally and

semantically,

and there

with the form

[N] [N-less] with the

odd,

as

Also, that

This analysis

it allows

perfectly straightforward structure

[N-less]A.

111).

seems to be no case

the structure

structure

would

of a compound

semantic structure [N]

[head] [strongness] seems be

realized

[strength], with level 1 suffix on the adjective. analysis

of these

items demonstrates

non-compositional word structure,

that

[head]

Selkirk's

once we

we can have a

unmotivated structures assigned to words.

as

allow

number of

25 We

have seen

above

ordering with the

in some

detail

adjacency condition has led

of non-compositionality of word has

a number

of

This is

cases of

results.

in which

semantic structure and the

how

There

which

exists

an

between the

lexical structure is maintained,

mismatch are

the position Kiparsky

2.2.

to the claim

the isomorphism

Kiparsky tries

given separate

(1982,

framework of Lexical Morphology. in

level

structure semantics,

undesirable

alternative approach

while the

that strict

1983)

accounts.

takes

in his

We have already discussed

to account

between lexical and semantic structures,

for the

which he calls the

'bracketing paradoxes', by proposing the following: allowing some limited recursion from

mismatch

first,

a syntactic level back

to word formation, in order to account for some instances of occurrence

of a

morphological

phrase in reanalysis

a

word;

under

second,

certain

framework

as Kiparsky's,

conditions

(See 2.2.)

accommodate level ordering violations. also mentioned that Pesetsky (1983),

postulating to

It was

working under the same

recently proposed

a mapping

of

lexical structure to logical form which involves a 'move~' transformation, as an alternative to Kiparsky's proposal for morphological Williams

and

reanalysis. Selkirk

in

Although holding

Pesetsky

lexical

follows

structure

and

logical form to be not isomorphic, he departs from others by attempting to

postulate a

between the two structures.

systematic mapping

relationship

26

Of the

two basic

ways to treat

the semantics

discussed above the latter approach to be more

taken by Kiparsky seems By

plausible and fruitful.

compositionality of word formation,

we

can

assuming the basic

we

possible structures for each word, mismatch

of words

can narrow down the

and even when there is a

investigate exactly

when

and

how

(and

perhaps under what circumstances) such a mismatch occurs, and postulate Another

some principles important issue

or rules here is

to

account for

the fact

them.

that among

the

cases of a mismatch we have some that involve a unit that is larger than a word-- not only It

is

interesting to

find

a compound but also a phrase. out

whether those

different in nature from word-internal cases.

cases If so,

are that

can be evidence for the strict dichotomy between the lexicon and syntax; if not, it can provide evidence against it. ~·

Typology of rules

3.1. Earlier attempts for rule typology So far formation

we have under the

been discussing assumption

various issues

that

the domain

of word of

word

formation is the word, and thus it is at least basically, if not completely,

separate from syntax

constituents that are larger than word. will discuss an approach whereby

whose domain is those In this section we

we look for the clustering

of properties among various processes, lexical or syntactic, and see grammar.

what the result can

tell as about their

status in

27 When Chomsky

(1972)

certain nominals syntax,

first

proposed that

should belong

he listed

to the

This was

the first

classify rules of

lexicon and

a few properties of the

morphological idiosyncracies and

process such as

in generative

grammar according to some however,

give justification for having two

separated

in

components

lexical redundancy

rules in

the

postulating the

Wasow,

(1)

changing,

The

local

properties

lexical rules

structure preserving, (3)

in

in

types of rules namely,

syntax and

them.

It

grammar,

transformations in

distinguish

criteria.

What Wasow did was

to try to

by

grammar to

that more serious and

detailed work was done on that attempt.

different

not to

semantic unpredictability.

attempt made

was not until Wasow (1977),

lexicon,

derivation of

are,

(2)

domain,

that

would

according

to

possibly category

(4)

ordered

before

transformations, (5) with idiosyncratic exceptions. 3.2. Rule vs. operation: Dowty's (1979) proposal Another important attempt of this type, this time between syntax/lexicon as made by Dowty.

well as

between syntax/morphology,

In trying to capture a realistic picture of

the distinction between morphology and syntax, proposes

that

'operation'.

we He

separate

lines

notions

morphological

Dowty (1979)

of

'rule'

productivity and

the essential

opposed to syntactic between

the

takes partial

unpredictability .as rules as

was

properties

rules,

of

and claims

operations

and

and

semantic lexical that the syntactic

28

operations may not always coincide this distinction, constrained

in

but sometimes cut across

He suggests that the operations might be some characteristic

for

ways;

instance,

morphological operations always give a fixed linear ordering of elements, the

while syntactic operations need not do so,

product

of

interrupted by points and

morphological

make some revision

are no longer

In any

be

test these case,

this

remarkable breakthrough because we

word or to ignore

also creates

later.

obliged to limit the domain

processes share

cannot

we will

syntactic operations.

approach of Dowty's is a

to the

operations

or

the fact that

properties with

a potential for

of lexical rules some morphological

syntactic processes.

explaining facts

It

observed in

typologically varied languages as we will discuss later. Table 1 shows this from Dowty

1979,

cross-classification of rules,

p.302,

except

for the symbols

taken

for each

category. {6)

TABLE 1. TYPOLOGY OF RULES AND OPERATIONS Syntactic rules

Syn.A.traditional syntactic ope. rules (PS-like and transformation-like)

Lexical rules B.rules forming lexical units of more than one word, e.g., Eng. V-Prt combinations.

Mor.C.inflectional morphology D.rules introducing deriv. Ope. 'derivational' morphology morphology, zero-derivation, when unrestricted and and compounding, partially semantically regular productive and less than (polysynthetic lang.) predictable semantically

29 The classes A syntactic rules

and D are what have been

often assumed as

and word formation rules

respectively.

B

and C represent the new class of processes which have been a problem in the previous approaches. units

larger

than

a

word

B consists of lexical which

non-compositional in meaning (idioms, best

regarded as

a

certain syntactic verb:

single

are or are

'word-like' constituent

taken advantage of).

under

passivized complex

The latter type of items have

often been given an account

those word

for instance)

rules (for instance,

(See Hornstein and

nonetheless

as an instance of 'reanalysis'.

Weinberg,

1980.)

formations which are

very similar

hav~

rules in their properties and

Class C

consists of to syntactic

been claimed to belong to

syntax by many people (see Sadock

1980 for such a case from

West Greenlandic), or else have been simply unanalyzable for those who maintain the lexicalist hypothesis. With

this four-way

distinction

rather than

previously

held two-way one, Dowty has given classes of processes which could not proper however,

previously be straightforwardly

place to that

belong to Dowty

lexicalist position syntactic rules;

nevertheless

that lexical

include

the set

C,

rules are

We should

note,

the

basic

ordered before

above,

rules of B

"since the domain of lexical

of basic expressions

expressions derived

for a

maintains

thus, in the chart (6)

and D always precede A and rules is

in grammar.

accounted

by

alone and

syntactic

values,

does not it

is

30 predicted that

in any

sentence in

which both

lexical and

syntactic rules are in evidence, the lexical rules must have applied before

any syntactic rules

have been

used,

hence

lexical rules are in a

sense 'intrinsically ordered' before

syntactic

• This theory

rules.

turns out

to predict

semantic limits on what the interpretation rule of a lexical rule can do" (Dowty 1979: 306). that it

is not

clear at all

We will see below, however, that this

claim is

true for

English; in later chapters we will examine numerous cases of phrasal affixation from Japanese. In the actual realization of the Din

(6),

Dowty

involve both

notes that

syntactic and

processes of types A to

a single

syntactic rule

morphological operations.

may He

gives as one example the subject and predicate construction; the PSR (NP +

VP)

belongs to class A and

the agreement on

the verb, if the subject is a third person, belongs to class C.

There is no problem of the ordering in this case, since

A and

C are

both syntactic

rules operating

on the

basic

expressions. 3.3. Interaction of processes What we will do in the rest of this section is to see how other combinations of

processes are realized in

actual construction in English. the

extent to

which Dowty's

forming an

By doing so we can find out program

with the

lexicalist

assumption works and where it breaks down; more specifically,

31 the

question

boils down

to

how

and where

the

boundary

between the lexical rules (B, D) and the syntactic rules (A, C)

is

relaxed in English.

We

will take up

the possible

combinations in turns. 3.3.1. A&B and C&D: iexicaliZatiOn One interaction between and C and

D is an unmarked

lexically are fed level a

a

by which what

word formation

and B,

is producd on a word

So,

feeds syntactic

for example, in the case of a compound noun

plural suffix,

a

inflection is added in C. lexical unit

one,

to syntactic processes.

product of lexical

word formation; with

processes belonging to A

is

formed in

On a phrase level,

formation is

according to the

compound

inserted to

as a lexical unit

and

a product of

syntactic processes

category they belong to;

verb-particle unit

D

for example,

of the V

a

category is

inserted accordingly to the syntactic structure. The other interaction of A and B, often called

'lexicalization',

and C and D is what is what is

namely,

produced

syntatically can become unanalyzable as

a unit on both word

Let us look at the

phrase level first.

and phrase level.

What is interesting about lexical seems

to always

sequence.

match

Idioms

produced by phrase is non-compositional

unit formation is that it

some part

generally have a

of

a surface structure of

structure rules even though (cf.

Zwicky

syntactic

1978);

as

a phrase

the meaning in kick

the

32

bucket (VP). to have

Similarly, a number of lexical units are found

a structure

of PP

or N' ,

(deal),

under-the-table

reanalyzed as

an Adj: (drugs),

over-the-counter

on-the-road (musicians),

matter-of-fact (attitude),

Other

not

lexical

units

constituents,

but

do

through

necessarily lexical

form

unit

interpreted as a word-size constituent;

etc.

syntactic

formation

are

for example, [take]

[advantage of x] --> [take advantage of] x. One interesting case involve

an

formation

output of

of

of lexical unit formation syntactic

complex adjectives

movement. with

It

'tough'

discussed in Nanni (1980), as the following: pattern,

an easy-to-take

formation

of

structure

where the

pattern

these

is easy

adjective

can be

medicine.

(for x)

rule

to sew

formed from

based

a

the

an easy-to-sew

on

that the

the

surface

has applied:

'this

y.' The

fact that

rather complex

where the surface sbject NP comes from

is

predicates

It is clear

adjectives is syntactic

seems to

this

structure

a PP is shown by the

following examples taken from advertisements. (7) Here, the easy-to-feel-pretty-in bibbed cotton flannel dress in beige or black, S-M $315. (Vogue, Oct. 1983) If you've just remembered another hard-to-buy-for person on your Christmas list • (WFMT radio, Chicago) The lexical rule of complex

adjective formation as proposed

by Nanni (1980) combines only an adjective and an infinitive form of a simple transitive verb, so it cannot generate (7).

33

It

seems

necessary that

some

kind

of reference

to

the

corresponding syntactic surface structure is necessary, from which a sequence of words are reanalyzed as a complex word. Complex

adjective

formation is

not

actually

strictly

limited to easy-type adjectives as Nanni (1980) assumes, but seems to exist with other adjective constructions which have been considered to be derived by a rule of object complement deletion (cf. examples:

Lasnik and Fiengo

1974).

See the following

3

(8) a. The girl is pretty to look at» --> a pretty-to-look-at girl b. The soup is delicious to taste)?! --> a delicious-to-taste soup c. The meal is ready to serve)?!' --> a ready-to-serve meal What these Adj.-to-V (part.) constructions have in common is that the subject the deleted they

NP (or the modified

or moved direct object

contrast

with

*ready-to-walk baby,

contrast

seems

to

be

parallel way:

are

more

corresponds to

and in

that point

cannot

form

related

like modifiers

to

a

*eager-to-succeed

*willing-to-talk person.

infinitives are subcategorized by they

NP,

Equi-adjectives which

complex adjectives in the man,

noun)

the

fact

that

This the

the Equi-adjectives while with

object

deletion

and

3 The adjective construction with ready is not mentioned in Lasnik and Fiengo (1974). Although it belongs to a different semantic group from pretty -type subjective evaluation adjectives, ready seems to be close to them rather than to easy -type adjectives, since it cannot have a sentential subject: *It is ready to serve the meal.

34 easy-type adjectives. difference in complement. how

this

Further, it may be attributed to the

the nature of

subject PRO of

an infinitival

See Chapter II, Section 1 for the discussion on contrast

may

be related

to

the

more

general

condition on word formation rules. Thus

we

can say

that

the

processes of

formation interact with syntax in two ways:

lexical one,

unit

a certain

part of surface structure

undergoes lexical unit formation;

two,

unit is

the

formed lexical

inserted to

a syntactic

As will be discussed

structure as a word-size constituent.

in later sections, one important principle that governs such surface restructuring

and reanalysis is the

basic argument

structure of predicates. On

the

word

level English

productive derivational of

such

processes

suffixation,

we

lexicalized items.

also

below.

One such case

arms, odds, pains, etc. semantically

irregular,

a

small class inflectional

small

class

of

is a plural suffix which

meaning of the word:

airs,

There are other plural forms which

transparent

but

are

morphologically

teeth marks, lice-infected.

problem for Dowty's

formed by

completely

which have been noted to occur inside a compound

(cf. Kiparsky 1982): be a

have

With

find only

has been incorporated into the

are

not

processes except for a

discussed can

does

a rule of

proposal,

class C,

after compound formation of D

because

This can

an inflection

which is supposed

to apply

must break into the compound.

35 If we

place the

class D,

on

Similarly,

morphologically irregular

the other hand,

this problem

lexicalized

the

inflection into

plural

does not arise.

(i.e.,

suffix

semantically irregular one), can be grouped as D, and indeed they also occur in compounds:

arms reduction, pains-taking,

odds-making. Another type of inflection

that interacts with processes

of D is the participle formation.

For instance,

of adjectives from passive participle lexical (cf.

derivation

has been argued to be

Wasow 1977) as opposed to the syntactic verbal

passivization.

A

few non-passive

derivation of an adjective as well:

participles fallen (tree),

allow sunken

{cheek), mistaken, etc. In

languages

with

more

processes this interaction has

been noted

productive

of C and D is

in Japanese

that

passive and causative verbs are

word

formation

more active.

It

productive formation

of

related to the transitivity

paradigm through lexicalization of some forms {cf. 1981, Miyagawa 1981, Kuroda 1982).

Jacobson

This point will be taken

up in Chapter IV. 3.3.2. g and Q: apparent case of WF involving phrases Some

complex

words in

English

which

appear to

be

a

product of word

formation applied to a

phrase actually are

derived by word

formation applied to a

lexical unit.

following are such examples:

The

un-[heard-of], un-[cared-for],

36 [laid-back]-ness (affixation);

[hand-out]V->N, [sit-in]V->N

(conversion); [pick-up] truck (compounding). Since

reanalysis on it

is

lexical

these

units

are

a sequence derivable by

natural

that

formation involving

they

give

a phrase.

unproductivity separates

appearance

the other

them from a

suffixation and compounding.

real case

adjective in

those that

word their

of phrasal

For instance, take the case of

every existing verb-particle combination but only

of

hand,

Note that not

un-[V- part.]A as exemplified by unheard-of.

form

lexical

syntactic formation,

the On

of

product

can

can appear in this

comfortably be

passive participle

form can.

example, unheard-of and *unheard-about.

used as Compare,

an for

In this sense this

type of word formation should not be confused with those that involve rules of A and C discussed in the following. 3.3.3. A & D vs. A~~: WF involving-phrasesGiven

the assumptions

processes A & C formation. processes of

here,

both

the combinations

and A & D involve phrase as

of

a unit of word

The difference between the two would be that the A & C have

no lexical exceptions

irregularities while those Dowty's version

of the

of A & D

do.

or semantic

Actually,

lexicalist assumption

under

that lexical

rules must precede syntactic rules, the combination A & D by which a syntactically generated phrase enters a lexical word formation should not

be possible;

on the

other hand,

the

37

combination A & C should be possible

since A & C

are both

syntactic. There are a small number of belonging to one of the two me at this point how they be that since

combinations;

syntactic

phrase

& D is

is

by

compositional (unless they undergo it

is

hard to

it is unclear to

should be distinguished.

the combination of A a

cases which can be viewed as

imagine

how

a

It may

logically impossible, nature

productive

and

lexical unit formation), word formation

rule

that

involves a phrase can be unproductive and non-compositional. I

will assume

process

here

without

much justification

of word formation involving a

~yntaetic

that

phraseis

any a

combination of A & C. What we are looking for, where one

of X

+ Y is

other a derivational constructions in

then,

a freely

affix.

English

is a construction X + Y generated phrase

There seems to be that

fits

this

and the

a class of description,

illustrated by the following examples: (9) warm-blooded, three-legged, tight-fisted, long-sleeved, deep-blue-eyed. As mentioned in

subsections 2.2.

and 2.4.,

examples like

these have been a problem for the strict lexicalist analysis of complex words.

This word formation process is productive

and compositional and can be schematized as: X+ed,where X is a N,•

means 'having X or

•It seems

characterized by the

that it is limited

presence of

to a noun modified

by one

38

X.'

It therefore seems plausible to consider this formation

as belonging to class C, often interacting with A to take in phrases as well (probably due to its productivity). The

class of

verbal

compounds

of English

exemplified

below resemble this adjective formation in some respects. (10) fast writer, hard worker, animal lover, party goer. three-time loser, early riser,

These words are quite productive formation

can be

schematized

and compositional, as:

X+er 'someone

as the who

is

characterized by the act/state of V (X,V)', in contrast with 'primary'

so-called lexicalized

to

compounds of

various degrees.

(10)

whose

meanings

Admitting

these

compounds,

as product

5

of syntactic

are verbal

word formation

yields some interesting results. As mentioned in 2.1.1, Roeper and Siegel (1978) a

lexical

According

transformation to

Wasow's

transformations

and

transformation'

itself

to

derive

(1977)

lexical is

verbal

compounds.

to

distinguish

proposal rules,

proposed

the

paradoxical.

term In

'lexical

fact,

the

formation of verbal compounds as we see it does not fit into Wasow's criteria--it Roeper and first sister

Siegel)

is structure-dependent in that the

to the verb

(as argued

first element must

in its subcategorization

word (or a compound as deep-blue) blue-eyed, *long blonde-haired.

thus,

*round

by

be the frame ,

(and)

5 0f course, many verbal compounds do undergo lexicalization especially because of their noun category.

39

which makes which

it syntactic,

makes

it

while it

lexical.

On the

classification of processes verbal to fit nicely the

word

is category-changing, other

hand,

suffixation)

our

compound formation seems

as a syntactic word formation.

formation feature

in

(i.e.,

It has both

category-changing

and the syntactic rule

by

feature that is due to

the involvement of V'-phrase. There are a couple of common peculiarities found with -ed Adjective formation and the verbal compound formation, which again separate them from seem to

primary compounds.

exhibit some degree

subpart can be modified,

Namely,

of transparency in

they

that their

inflected with comparative ending,

and pronominalized by one, as shown below. (11) a. b. c. d.

John is warmer-hearted than Harry is. This figure is exactly/more than five-cornered. He is the earliest riser in the house. I want to be a beautiful dancer, not a poor one.

Semantically they also form a complex

seem to share the

feature that they

predicate as an adjective derived

category or as a nominal derived

from a N'

from a V' category.

(See

Chapter II for more discussion of verbal compounds.} In

Japanese it

important Chapter IV

category

is

found that

in various

and Chapter V will

phrasal suffixation in Japanese, this A & c combination.

verb

word

phrase

formation

be devoted to

is a

very

processes.

discussion of

which presumably belong to

40 3.3.4,

~ and~

The last

combination of processes involves

formation and productive word

formation.

word formation are few in English,

lexical unit

Since productive

and they do not seem to

feed lexical unit formation, the only notable interaction is the lexical unit undergoing saw that lexical

productive word formation.

units undergo lexical word

word-sized unit.

Inflectional endings,

We

formation as a which belong

to

syntactic word formation of class C, are found not to attach to them as a whole but only care of] (Passive), nominal).

So we

characteristics

to the head word:

the [turning on] of may say

with

that

regard

morphological operations:

[takgn

the light (action

a lexical to

cf.

the

unit has two

it is transparent

types

dual of

to inflection

and opaque to lexical word formation. 3.3.5. Further remarks Now that

we have seen some

examples of how

the various

processes are realized in English, let us elaborate a little on our observations.

In the rule typology presented above,

lexical unit formation (B)

and syntactic word formation (C)

are the deviation from more 'traditional' classes of A and D where the

types of

speaking,

lexical unit formation involves a 'reanalysis' of

a part

of surface

rules and

operations match.

syntactic sequence

syntactic word formation

into a

involves a kind of

word,

Roughly

while

'amalgam' of a

41 phrase structure and a word structure.

That is why lexical

unit formation never 'creates' a structure,• while syntactic word formation is capable of

creating a structure that does

not match the phrase structure Chapter IV for

rules of the language.

such structure building and

(See

its consequence

in Japanese.) Formation of a lexical unit out not take lexical

place randomly. units in

intransitive

(1980) call

It seems

English form

predicate.

subcategorization

of

In

that the

either a an

examples of

transitive or

intriguing

complex verbs,

study on

Carlson

and

an the

Roeper

noted that rule-created complex verbs have what they 'unmarked'

subcategorization,

transitive or an intransitive more marked ditransitive verbs PP.

of a group of words does

verb~

that

of

a

rather than that of the

or verbs subcategorizing for

They found this generalization

verbs of different origin;

namely,

to hold across complex

syntactic or lexical, phrasal or

morphological. On

the other

hand,

there

seems

to be

no reason

lexical units to be limited to those involving verbs. (1977)

Wasow

also questions the assumption that his lexical rules

involve only verbs.

We have seen that

can also be a lexical unit. point

for

whether

lexical

unit

complex adjectives

It is an open question at this formation

is

"Except for the case of a frozen syntactic earlier stage, such as 'battle royal'.

limited

to

phrase of an

42 argument-taking 'predicates'. 3.4. Cliticization Although it is not often viewed word

formation,

relevant to

the

process of

cliticization

our discussion here,

characteristics especially our

as part of morphology or

classification,

for it has

to the processes

that

is,

is

quite

some similar of class

syntactic

C in

rules

with

morphological operation. Clitics can be described as lean

for support

Greek word their

(the term

for 'leaning')

construction.'

well-known examples ('je l'ai king of

vu' Fr.),

detail, levels

'clitic' on a

originates from

neighboring full

(Matthews

1974:

168)

of clitics are Romance the

France's hat'),

('arma virumque').

'unaccented words which must

English possessive and

the Latin

As discussed in

the

word in

Among

the

pronoun clitics marker

('the

conjunctive marker

Zwicky (1977)

in some

clitics can present a number of problems on various of

linguistic

phonological,

and

analysis;

syntactic,

semantic,

morphological.

Since it is

beyond the

scope of this section to discuss associated with

cliticization,

the full range of problems we will

only touch

on the

aspects relevant to the framework of rule typology we pursue here.

The basic problem that cliticization presents in our

discussion is the following: to a

syntactic phrase,

although clitics often attach

they are

joined only to a

part of

43 that string.

So, in the case of English possessive marker,

for instance,

although the marker

applies to the whole NP,

'the king of France' in the example above, to the last

word.

Thus if we were to

phonological word boundary as marking

it attaches only

strictly follow the the lexical unit,

it

would result in a mismatch between the semantic constituency and the syntactic constituency; the

'bracketing

discussed in extends to

paradox'

2.2.

of

above,

a more

a problem quite parallel to level-ordered

except that

morphology

this present

significant boundary

of syntax

case

and the

lexicon. It is rather natural,

then,

confronted with this problem to

treat

cliticization

morphology and

to place

that a number of linguists

of cliticization have proposed

separately it so

from

that the

the

rest

of

surface syntactic

strings can feed cliticization, which is then viewed as part of phonological suggested

that

readjustment rules. cliticization

Aronoff (1976)

should

be

grouped

inflection (which belongs to syntax in his view) to derivation. on

the other

has with

as opposed

Zwicky (1982) and Zwicky and Pullum (1983), hand,

propose

that cliticization

makes

a

completely separate component from inflection or derivation, and maintain the

that cliticization component is

syntactic component.

Zwicky

ordered after

and Pullum's

Interface

Model (see Zwicky 1982) predicts that separate components of inflectional morphology,

syntax,

and cliticization are all

autonomous and do not overlap nor interact freely.

44 There

is

some

comfortably

claim

evidence,

that cliticization

phonological phenomena

and dismiss

discussion of morphology. an inflectional Pullum (1983) historically

however,

suggest that it a

counterpart), affix can

a

cannot

postsyntactic

it altogether

to a clitic,

from the

Zwicky and

may very well have developed

'simple'

clitic

readjusted form of a free form). could also look

is

we

After arguing that English n't is

affix as opposed

from

that

(a

phonologically

They also suggest that it

like a 'special' clitic

(without free form

and that a special clitic and an inflectional

be very

closely related.

Thus although

their

contention is to separate clitics and inflection as discrete phenomena,

the

boundary cannot

always be

made clear-cut.

This point is elaborated on by Sadock (19B3a,b), that on both be a

ends of derivation and

gradual transition

He takes transition

data from (cf.

Greenlandic to

Sadock show the

inflection there could

from cliticization

English to

who claims

show the

1983a),

and

to affixation.

inflection/clitic data

from

West

derivation/clitic transition

(cf.

Sadock 1983b).

There is no space here to discuss his data,

but

will

instead we

briefly see

the

implication of

the

existence of such facts to our scheme of rule typology. First, it is easy to see that the transition from clitics to

inflectional affixes

phrasal operation rules (namely,

resides

in

the boundary

and morphological operation A and

C in the

chart).

In

between

in syntactic other words,

45

although cliticization on the

degree

of

operation is)

the whole belongs to

'wordhood' (i.e., can vary

how

category c,

morphological

diachronically or

synchronically.

Further,

the more morphological the operation is,

there is

likelihood of

hence of the of

our

clitics showing

Sadock's (1983b)

the more

idiosyncracies and

process become part of lexical

typology).'

the

rules (class D

discussion

of

West

Greenlandic noun incorporation suffixes is most revealing in this connection.

He demonstrates that in West Greenlandic,

a language with extremely well-developed morphology, various noun incorporation suffixes can be to their behaviors and properties, smooth

transition

from

put on a scale according which

shows there is a

free-word-like

derivational bound morphemes.

What this

suffixes

to

means in our rule

typology scheme is that in a language where much syntax must be carried

out with morphological

operation,

the

unit of

word must accommodate various degrees of lexicalization, and hence in such a case the

boundary between C and D processes

(syntax/lexicon boundary of morphological operation) must be particularly gradual.

Note here that both

derivation have been shown to phrasal operation processes; with our contention

inflection and

exhibit fuzzy boundaries with this observation is consistent

that inflection and derivation

are not

'Such idiosyncratic behaviors are part of what Zwicky and Pullum (1983) list as criteria for distinguishing affixes from clitics; namely, affixes are highly selective on hosts, their paradigm can have gaps, and they show phonological as well as semantic irregularities.

46

treated as two distinct components under the category of do not

mean to

but are grouped together

morphological operation,

ignore the

difference that

although we

exists between

them, especially that of relative ordering. We have seen then that cliticization, a

category C

may not

process,

phrasal syntactic rules

always

of A or from

rules of D.

There is

can present:

if we assume

a

where

from

after syntax?

phrase

chapters when

(A and

those in which We

It

as

a

how can we distinguish unit

undergoes

C combination

we discuss

a phrasal clitic

will be

many potential

is attached in the later

cases of

as well as a few

some

discussed in

will take up this question

suffixation from Japanese English.

morphological lexical

are phonological readjustment rules

morphological operation 3.3.3.)

from

that the cliticization processes

fed by surface syntactic structures, cases

be distinct

a further problem that cliticization

(at least some of them)

the

which is basically

phrasal

such cases from

argued that although

we do

need to

attribute some of them to postsyntactic phonology, it is not plausible to treat all of them that way. Some

observations

relevant here. typology,

on

language

Sapir (1921),

suggests

that the

typology

seem

to

be

in discussing morphological popular classificatory

terms

such as 'analytic', 'agglutinative', and 'polysynthetic' are not properly understood, 'agglutinative'

as

being

but that we should view the notion in

opposition

to

the

notion

47

'fusional',

thus referring to the

mode of morphology,

and

the notion 'analytic' should be considered to be relative in degree to 'synthetic' and 'polysynthetic'. relevant to

the discussion of

opposition.

Sapir

'juxtaposition')

What seems to be

cliticization is

says

that

the former

'agglutination'

(or

as opposed to 'fusion' is characterized by

directness and mechanicalness of affixation; in other words, low degree

of idiosyncracy in phonology,

semantics.

fusional languages such as Greenlandic

above. this

and

Note that this is part of the criteria used for

distinguishing cliticization

West

morphology,

from affixation

in relatively

(analytic~fusional)

English

(polysynthetic-fusional)

as

and

discussed

It is therefore very important that we bear in mind partial

similarity

between

agglutination

and

cliticization when we look at an agglutinative language such as Japanese. suffixation

In the later chapters when we discuss phrasal in

Japanese

we

will have

difference between processes that

this

partial

parallelism

look

for

the

are clearly postsyntactic

and those that interact with syntax. how

to

We will also consider

between

agglutination

and

cliticization affect the syntax/lexicon interaction. 4_. Lexicalism in Japanese Syntax Being

an

agglutinative language,

Japanese

exhibits

rather complex morphology, especially in verbal suffixation: for instance,

the following is an

multiple suffixes:

example of a

verb with

48

(12) hatarak-ase-rare-ta-gari-sugi-ru. work-cause-pass-want-act-excessively-present 'act excessively like (he) wants to be caused to work' In the generative

approach to Japanese syntax suffixes are

a number of verbal

in

analyzed as

the past,

higher verbs

taking a clausal complement (cf. Kuroda 1965, Kuno 1973, and Inoue 1976).

In other words,

derivations such and

case

as passive

marking

for the purpose of syntactic and causative

assignments,

transformations

verbal

suffixation

is

de-agglutinated and each suffix is treated as a higher verb. Recently, approach

however, and

the

theory (which

with the

influence of

autonomy thesis

was discussed in

in

the lexicalist

general

Section 1),

linguistic

some attempts

have been made to treat those verbal suffixes in the lexicon and to view

the rules like passive

as lexical rules rather Farmer

1980,

than syntactic transformations (cf.

Miyagawa

1980).

hypothesis it is significant that rare are bound cannot be a emphasized parallel

and causative formation

morphemes,

Under

the

suffixes such as sase and

since it follows

then that they

unit in the syntactic operations. that

sets

complex of

verbs V-sase

case markings

lexicalist

to

and simple

It is also v-rare

assign

verbs.

For

instance, V(intr.)-sase assigns transitive case markings and V(tr.)-sase assigns

ditransitive case

markings.

In

sum,

verbal suffixation is treated as a lexical rule that changes the argument verb.

structure and

There has

the case

markings of

not been much rebuttal

the base

to the lexicalist

49 approach to camp'

Japanese syntax

except for

approach

Kuroda

seems to

criticizes

from the

(198la,

be gaining

the lexicalist

derivational morphology

198lb),

support.

approach

and

Kuroda

by

this

that

an independent

should be autonomous from

sense that the scope of

a syntactic rule

should not be limited by the unit in morphology, i.e., word.

new

(198la,b)

pointing out

should not be given

theoretical status because syntax morphology in the

'transformationalist

to a

He attempts to illustrate this point by showing that

the causative

morpheme sase is

morpheme with the sentences

a stem

as well as

a bound

where it appears independently;

for instance: (13) Taroo ga Jiroo ni utai mo sase-ta. NOM DAT sing also cause-PAST 'Taroo let Jiroo sing as well.' The

emphatic marker

causative morpheme. so

clear-cut,

mo

separates the

verb and

Unfortunately,· this argument for there

this occurrence of sase is

actually suru 'do'

the passive morpheme rare or of suru is

the

is not

is evidence that suggests that

turns out to be s-ase)i namely,

the presence

main

-(~)ase(which

in the parallel cases with

the desiderative morpheme tai,

unambiguously required,

as shown

below. (14) a. Taroo wa Jiroo ni utai mo s-are I *rare-ta. 'Taroo was troubled also by Jiroo's singing.' b. Boku wa utai mo si-tai I *tai. 'I want to sing as well (among other things).'

50 Kuroda (198la, 112) himself notes this possibility that it is not conclusive as as a

to whether sase is really a

bound morpheme.

8

Nevertheless,

theoretical point

what belongs to syntax cannot

Kuroda makes is a valid one: be by assumption

the

stem as well

limited by the morphological

unit.

This

point is especially important for an agglutinative language, where

the

morphological

concatenation can

be

loose

and

transparent. We

will

briefly

elaborate

here

mentioned and discuss what seems

on

this

point

to be problematic with the

lexicalist approach to Japanese 'syntax'.

One assumption to

be questioned is that argument structure (or case frame) uniquely associated with the word unit; are assigned case frames that structure while below,

concatenation different in

and

there

are

i.e., complex verbs

cases

and

where

causative formation

and

periphrastic

concatenation

are

sometimes even

in their

regularity and transparency of application. productive

As exemplified

are not.

agglutinative

their function,

is

are realized in the syntactic

bound morphemes

however,

just

not

First, take the

the periphrastic

V-te

morau construction. (15) a. Taroo ga Jiroo ni uta o utaw-ase-ru. NOM DATsongACCsing-cause-PRES 'Taroo causes Jiroo to sing a song.'

8 In regard to this point, see Sugioka (1982) for the discussion on suru, which exhibits varying status from a bound morpheme to a free word depending on the environment.

51 b. Taroo ga Jiroo ni uta o utat-te mora-u. NOM DATsongACC sing receive 'T. receives the favor of J.'s singing a song.' In (15b)

-te is a continuative marker on the verb and morau

is a free word which can be used independently: (16) Taroo ga Jiroo ni hon o morau. NOM DAT bookACC receive 'Taroo receives a book from Jiroo.' Nevertheless, same way case

-te morau

in the

assignment of

markings to

NP's,

as

Further,

productivity. two words,

and -sase function in

they

the argument

well as

exactly the structure and

exhibit the

although utatte

cannot be separated by

complete

morau consists of an adverb,

which,

naturally, is the case with utaw-ase. (17) *Taroo ga Jiroo ni uta o utatte kinoo morat-ta. NOM DATsongACC sing yesterday receive-PAST 'Taroo had Jiroo sing a song yesterday.' (cf. Taroo ga Jiroo ni hon o kinoo morat-ta.) Should -te morau be treated also as a suffix to capture this parallelism?

This move not

assumption of

separating phrase unit

also would be

problematic in view of the fact

endings

do

behave

process (which triggered by

only undermines the lexicalist

differently

can be

under

considered as

the feature

process applies basically

on the

from word

the

but

that the two honorification

a kind

of inflection

subject NP),

to the head word

unit,

since this

of the verbal:

9

9 Although (18a,b) show a clear contrast, the honorification process of a complex verb is a complex matter, as discussed in Kuno (1983).

52 (18) a. Sensei ga Taroo ni uta o o-utaw-ase ni nat-ta. professorNOM DATsongACC sing-cause-HONOR-PAST 'The professor made Taro sing a song.' b. Sensei ga Taroo ni uta o utat-te o-morai ni nat-ta. prof. NOM DATsongACC sing receive-HONOR-PAST 'The professor received T's favor of singing a song.' (cf. *o-utat-te morai ni nat-ta) It is clear

from the preceding discussion that

word is less

obscure under certain processes

recognized in

others.

Our

rule typology

the unit of while clearly

scheme captures

this in the following way: (19)

syntactic rules

lexical rules

------------------------------1----------------------------phrasal

I

V-te morau

---------------~--------------1----------------------------

morphol.

V-sase

The similarity

and the difference

follow from the assumptions that structure

is

a

honorification

domain is

an

of

between the

the assignment of argument

the

syntactic

operation

sensitive

boundary.

Also

the characterization

with

fact

that

the

v-~

incorporated into a compound much

two endings

can

rules to

above is be

the

while word

consistent

lexicalized

and

more frequently than V-te

morau, because of its morphological nature. Another such parallelism can be found with 'raising'-type constructions

but

phrasal to suffixal.

with

a

more

gradual

transition

from

Note the following examples, 10

10 Examples of so-called 'raising' (20a,b) have been given much attention but their relationship to (20c,d) have not been discussed in the literature.

53 (20) a. Taroo ga [zinsei ga munasii] to omou. NOM life NOM empty COMP feel 'Taroo feels that life is empty.' b. Taroo ga zinsei o munasii to omou. NOM life ACC empty COMP feel 'Taroo considers life to be empty.' c. Taroo ga zinsei o munasiku omou. NOM life ACC empty feel 'Taroo considers life empty.' d. Taroo ga zinsei o munasi-garu. NOM life ACC empty-feel 'Taroo considers life empty.' Although (20a-d) vary

from

are very close in what they express,

completely

uni-clausal with in

between

(20a)

the 'lower' predicate

are

complementizer

hi-clausal

uni-clausal

(20b)

complementizer (20c).

and

case

to

they

completely

incorporated (20d);

frame

uni-clausal case

with

retained

frame with

no

We will leave the detailed discussion

especially of the suffix garu to Chapter III,

but only note

here that the transitive case frame of (20b-d) should not be given a completely separate accouont, one hand and in the lexicon (20d) also be noted

main

on the other.

that there is no lexical

constructions-like (20a) the

verb

and

the

while in 'lower'

restricted,

and in those like (20d)

be suffixed

by garu

feelings.

in syntax (20b,c)

are only

It should

idiosyncracy in the

those like (20b,c) predicate

are

both rather

the adjective that can

those of

emotional/physical

(See Chapter III for the details.)

observe here that the lexical

on

It suffices to

idiosyncracy does not clearly

54 divide the word level (20d)

and the phrasal level (20a-c),

nor does the case frame. We have provided evidence

that the lexicalist assumption

that the unit

of word is uniquely

structure

too simplistic

data.

11

is

in

a phrase.

above

some

framework,

We have

consequences

the

discussion

For

situation

an

of

this

of

a

is more

examination

of such

of

morphological

ways,

there

operations

Section 2

the

general

of

be weakened.

such

cases

see Chapter II, like

we find

They

section the

many cases

of

syntactic phrases.

merit

and morphology

a careful do

are

properties

(as

opposed

We to the study

interact in particular

to

in

Japanese

the following chapters

cases.

although syntax

in

claim must

language

attaching to

a large part of

claim

handful

interesting and

will devote

number

actual

English it is mainly with

that this

agglutinative

derivational affixes

because

the

already discussed in

derivational morphology of English 1.)

face of

attaches only to a word and

and observed that for

inflectional morphology (For

the

The other assumption that must be empirically tested

is the claim that a bound affix not to

associated with argument

the

a to

phrasal

operations) or to syntactic rules (as opposed to the lexical rules) that constrain the possible types of interaction.

As

11 See Carlson and Roeper (1981) for the discussion of the parallelism between prefixed verbs and 'syntactically generated complex verbs' in English: e.g., unroll the rug vs. roll the rug out; mistreat him vs. treat him poorly.

55 a

conclusion to

this

departing

from the

observing

a

section,

simple

broader

it

is

lexicalist

range

of

suggested that

by

assumptions and

by

processes

belonging

to

derivational morphology (i.e., not just those that have been accounted formation),

for

in

syntax

we will be able to

such

as

passive/causative

approach closer to the true

picture of how morphology and syntax interact in Japanese.

CHAPTER II DEVERBAL NOMINALS AND COMPOUNDS This chapter examines some aspects of verbal compounds in English

and

concerned

Japanese.

about

is

What the

will

relationship

structure of a verb and verbal will look at

we

rule

between

be

argument

compounds.

In Section 1 we

various frameworks that have

been proposed to

capture certain generalizations about discuss the

particularly

implication of

typology.

Section

this relationship and

those claims 2

will

to our

scheme of

examine Japanese

verbal

compounds with the

focus on how abstract

nominals are used

in such compounds

to reflect the argument

structure of the

source verb, and contrast them to the case of English verbal compounds.

Section 3 will discuss

phonological process

some obervations on the

'rendaku' that frequently

takes place

in the compounds of Japanese. 1. English verbal compounds and argument structure 1.1. First Sister Principle English verbal compounds, as opposed to primary (or root} compounds,

can be

defined with the following

they are marked by the and -ed,

morphological markings,

and they have a

verb base head (cf.

56

two points: -ing,

-er,

Roeper and

57

Siegel 1978).

The following are

some examples

of verbal

compounds. (1) truck driver, party goer, fast mover, good looker, house cleaning, eye catching, nice sounding, hand made, teacher trained, fresh baked, well written Some generalizations about verbal compounds and the problems they present to mentioned in

certain frameworks of morphology

Chapter I.

them in greater

In

this section we

detail and propose an

have been

will discuss

alternative approach

to verbal compounds. One salient generalization observed (1978)

is expressed in their

all verbal compounds

by Roeper and Siegel

First Sister Principle (FSP):

are formed by incorporation

To exemplify how FSP

in first sister position of the verb. works,

we can look at the first example in (1)

truck driver,

truck is a direct

sister of drive. drive

has

an

adverbial phrase (at) same reason,

mentioned

and

driver,

where

therefore

the first sister.

in

*guick-maker (cf.

Chapter I

(2.5)

approach to morphology proposed by

the

For the

in

the

noise-maker). level-ordered

Allen (1978)

FSP cannot

be a principle constraining word formation rules, postulates that compounding affixation,

In

an unambiguously transitive verb cannot form a

compound with an adverb: As

meaning

night is

above.

object and hence the first

Contrast this with night intransitive

of a word

and her

compounding 'blind'

must be on a

higher level than

Adjacency Condition makes the to the

since she

derivational history

rule of (that is,

58 deverbal

nature)

proposes that

of

the

FSP can

operate as

which is not subject to led to contend

compounded

items.

Allen

an interpretation

the Adjacency Condition.

that verbal compounds and

are the same in formal

nature,

thus rule,

Allen is

primary compounds

which claim is subsequently

maintained by others (Selkirk 1982, Lieber 1983). 1.2. Verbal compounds vs. primary compounds We will argue here that the analyses of English compounds which

do

not

formally distinguish

primary compounds prediction. (1978)

are inadequate and

make a

compounds

and

wrong factual

First, let us recapitulate Roeper and Siegel's

claim

that verbal

predictable in

meaning,

though a number of their

verbal

compounds are and

compositional and

extremely productive.

Even

verbal compounds undergo lexicalization,

basic meaning

is

always

predictable.

Thus

truck

driver differs from bike rider in carrying the meaning of it being

a

"someone

profession, who V's

but

X

the

basic

(professionally,

meaning of

both

habitually,

is

etc.)"

Following this pattern, one can immediately make a number of new

compounds.

productivity compounds

is

apart

Such typically from

transparency

in

syntactic,

primary

compounds.

and 1

meaning sets For

and verbal

instance,

1 This does not mean that the meaning of primary compounds is totally unpredictable. On the contrary, in a large number of cases their meaning can be systematically accounted for, and the reader is referred to Levi (1978) for a detailed study. Our contention, however, is that the meaning of verbal compounds can always be directly be

59 Downing (1977)

discusses the result

conducted in order

to test the creation

of novel N + N (primary)

compounds,

constraints on them cannot be semantic

or

syntactic derived.

interpretation

is of

based varying

and interpretation

and concludes that the

characterized in terms of the

structures

(conceivably)

relationships

of the experiments she

from

Rather, on

which

she

says,

permanent

semantic

they

are their

non-predictable

types,

and

"their

tendencies indicate more about the process of categorization thari they do about derivational processes."

She

constraints on the compound

further concludes

it is

the size

of the

temporal and spatial range of speech situations in which the compound

is used

to

name an

entity

that constrains

the

speaker in his use of the compound. Consequently, the analysis which does not distinguish the derivation of verbal compounds

and primary compounds cannot

capture this difference between them. an analysis, however,

The adherents of such

might argue back to this claim saying

that incorporating FSP as an interpretation rule can capture the same generalization.

For Allen,

this was necessary to

keep her scheme of level-ordered morphology, but it also led to

the

divorce

of

the

morphological structure. verbal compounds

semantic

structure

Selkirk (1982)

and primary compounds

from

the

claims that both are derived

obtained from their derivational processes, primary compounds.

as

by the

opposed to

60 same set She

of WSR's,

gives

hence

what roughly

amounts

additional observations stated in terms of the verb (ala LFG). be satisfied

they are

formally non-distinct.

to

FSP

in the

on the verbal compounds.

form

of

They are

2

functional argument structure of the

(a) The subject of a lexical item may not

in compound structure.

All non-subject

(b)

arguments must be satisfied within that compound immediately dominating

the

henceforth

Observation (a),

head.

"non-subject

call

compounds)" and

discuss in

which (on

condition

greater detail

later,

particularly emphasized by Roeper and

Siegel,

based

the

on

the

Standard

subcategorization

Theory

Observation (b), general

at

least)

phrase structure; category

first-ordered

does

on

namely, i

X

projection

word

that (a)

will verbal

is

must

be

of

~.

verb

which

include

and (b)

(in

subject.

is a case of a more

structure as

satisfied We

not

since FSP is

well

all non-subject arguments

problems with this observation later, to observe

not

according to Selkirk,

principle holding

lexical

of

we

will

as of a

within

the

discuss

the

but it suffices here

together can

eliminate the

2 Selkirk actually gives a different definition from Roeper and Siegel's for verbal compounds; namely, she includes basically all deverbal nouns and adjectives (regardless of their endings) as an eligible head of the verbal compound, but she limits the non-head element to being a subcategorized (i.e., non-oblique) argument. See the discussion below that her definition fails to capture the salient class of compounds and that one of her observations on verbal compounds (as she defines them) does not hold true.

61 majority of

the violations of FSP

as far as

Selkirk's own

notion of verbal compounds (see fn. 2) is concerned. however,

that both (a)

verbal compounds

and (b)

and they

Note,

are merely observations on

do not

follow in

any way

from

their derivational processes. Let us take a brief look at another analysis of compounds where no primary

distinction is made compounds.

morphemes

(stems

In Lieber's

(1983)

affixes)

have

and

containing information about argument structure,

between verbal

and so

compounds and

framework, lexical

subcategorization, on.

They

all

entries semantics,

are inserted

onto

lexical structure trees according to their properties, and a set

of feature

percolation conventions

ensure the

features to be transmitted to the lexical head. that we do compounds

not need to make any and

occurrence can

primary

She claims

distinction between verbal

compounds,

be predicted

proper

but

by Argument

their

possible

Linking Principle

(ALP), which says that argument-taking words (V and P) 'link' all

internal arguments

in the

The reader is referred to Lieber how

this mechanism

impossible compounds

is

supposed

of English.

(1983)

compound structure.

3

for the details of

to predict We

must

possible

will only

and

note here

3 Lieber adopts Williams' (1981) notion of argument structure, where subject of the predication is called 'external argument' and the other subcategorized arguments are called 'internal argument'. By restricting ALP to internal arguments she automatically incorporates the 'no-subject condition.'

62 that

even

though

ALP

applies

to

those

that

involve

argument-taking V or P, her compound structure always allows insta~ce,

another structure where it does not apply; for

for

truck driver: ( 2)

b.

& .

N

~

N N -t..-u.c.~~ V

N

d"'\"e.

e.r

In (2a) ALP applies and gives the prediction similar to that of FSP

(si~ce

the

object truck), given an

but

ve~b's

internal

in (2b)

argument is

ALP does not apply,

arbitrary interpretation (as a t~at

It then seems

the direct so it is

primary compound).

ALP does not really cover

any more than

FSP does, considering that it applies 'vacuously' to primary compounds;

although

Lieber

claims

it

captures

the

generalization covering both types of compounds. Now that we compounds and

have reviewed three approaches primary compounds

are not

where verbal

distinguished and

the generalization of FSP is achieved by other means, let us go back

to the

question of whether

their analyses

can be

claimed to be equivalent or superior to Roeper and Siegel's. We will argue here that they prediction

where these

cannot.

analyses

go

There is one factual wrong;

namely,

the

distribution of VP adverbs in the compound structure.

Since

a VP

it is

adverb can

allowed to

e~ter

occur as first a

ve~bal

sister to

a verb,

compound in Roeper

and Siegel's

63

analysis (hard worker, fast mover, early riser, etc.).

On

the other hand, they cannot give an adequate account for the absence of =lY adverbs in nominal compounds, such as in beautiful(21Y)

dancer.

This

problem can

adverbial marker lY

treating the

as an

be solved

by

inflectional affix

which consequently never enters derivational morphology,

as

argued for

be

in Sugioka

and Lehr

pointed out that a VP adverb but never

(1983).

It

can then

may appear in verbal compounds

in primary compounds;

it is because

VP adverbs

enter a compound as a modifier to the verb from which a noun or an adjective is derived.

For instance, the compound hard

worker can be formed from [[work]

[hard]]V' with hard as an

adverb, but the primary compound [hard] [worker] cannot have hard

as

an adverb

but

only

interpretation it can have. in

the

approaches

compounds

are

not

where

Similarly,

a

verbal

examples include etc.

where the

compounds

and

category

primary

adverb

is

For instance, Selkirk does not have she includes in her examples

non-Jx adverb

although Lieber

lexical categories

whatever

It is then not surprising that

adverbs in her WSR even though with

adjective,

distinguished the

mysteriously obscured.

compounds

as an

such

as

hard

working.

does not include adverb

that can enter compound quick-considering,

in the

structure,

her

carefully-considering,

first element is labeled

a Manner adverb.'

'It is argued in Sugioka and Lehr (1983) that all the apparent cases of lY-adverbs in a compound, such as carefully considering is an instance of a phrase, ADV + v-ing/ed, and not a compound.

64 This fact

about the

accounted for the verbal

distribution of

in an analysis

compound head is

VP adverbs

cannot be

where the deverbal

source of

not recognized.

supports the

analysis which distinguishes

from primary

compounds by deriving

It therefore

5

verbal compounds

the former from

a verb

and its argument. 1.3. An alternative analysis of verbal compounds Having

argued that

Siegel (1978)

the

basic

is correct,

let us

analysis is actually flawless. which

have

been

mentioned above.

given

There almost

Roeper

examine whether

and their

are a few criticisms all

of

those

papers

One of the criticisms concerns the notion

of 'subcategorization' (1978).

by

approach of

as it appears

in Roeper

and Siegel

They consider the subcategorization frame to appear

as the following (R & S's (122)): (3) Verb [DO] [Adv] [Inst] [Agent]

5 There are problematic examples with S-adverbs: e.g., 'possible/ certain/accidental winner'. If they are derived from '(he will) possibly win', it suggests that the input constituent (V, X) can be a V' with a uniquely surface structure origin (presumably possibly is a sister to S in the base structure), which in turn suggests that the 'feedback' is from a surface syntactic structure to the word formation. However, this does not seem to be a genuine case of verbal compound, since FSP can be violated: e.g., possible winner of the Nobel Prize. This type of compounds remains a problem for all the accounts of compound formation including FSP and ALP.

65 In addition to those in {3) such

arguments

as

it is clear that they also need

[Locative]

{home-cooked),

{church-goer), [Time] {night-driving), and so on. to the commonly perceived view, includes only those

however,

[Goal] According

subcategorization

arguments that are obligatory

verb, and not those optional elements such as time, and so on. from

her

with the manner,

Selkirk, therefore, excludes optional arguments Functional

Argument

Structure

definition of verbal compounds {cf. fn.

and

2).

alters

her

That does not

seem to be the right move, however, since the generalization captured by FSP extends over adverbial arguments as well, as discussed at the beginning of truck driver although

this section with the example

L night driver.

she also

complication on

In the case of Lieber {1983),

criticizes

Roeper

the subcategorization

and Siegel frame,

and

for

the

her ALP

refers only to the internal arguments, it is also crucial to her account whether certain

semantic

a certain item can be argument

or not.

interpreted as a

She

excludes,

instance, ballad-picked as a possible compound, first element oblique) salient

cannot be

argument.

interpreted as

(i.e.,

It is clear that in order to capture a

generalization about

as far

because the

a semantic

verbal compounds

include oblique argument of the verb as well. hand,

for

as the lexical rules

we need

On the other

are concerned,

arguments do not generally play a role.

to

oblique

66 Let us propose an alternative

view:

what we are dealing

with here is not what is subcategorized by a verb but rather what can from a

occur under a

V'.

possible expansion of

constitute a well-formed V', verbal compound. more tricky.

Verbal compounds V',

V + X.

are derived

If this V

+ X

[X V]-er/-ing is a well-formed

The case of passive participle is a little Actually,

the

participle compound (ex.

first element

of a

passive

teacher-trained) is a first sister

not to the main verb but to the passivized verb (trained (by a) teacher).

It is not the affixation of -ed that form this

complex adjective but

rather

a~-derivation.

Let

us list

the verbal compound rules: (4) a. [V X]V' + er ---> [X V-er] N b. [V X)V' + ing ---> [X V-ing] N, A c. [Vpass. X]V' + )0 ---> [X Vpass.] A Note

that

they

turn

out

to

be

exactly

parallel

to

nominalization and adjectivization of a simple V: (5) a. V + er ---> N (ex. writer) b. V + ing ---> N (ex. painting), A (ex. exciting) c. Vpass.

+~

---> A (ex. surprised)

Another major criticism to Roeper

and Siegel's analysis has

been that they need to postulate a separate set of rules for deverbal nominals and adjectives besides the verbal compound formation rule, of affixes.

although they involve

In our

exactly the same set

alternative analysis,

formation can be perceived as an

verbal compound

extension (in the sense it

67 involves

V')

of

nominalization and

adjectivization of

a

verb, thus eliminating this redundancy.• From semantic considerations (4)

and (5)

are related.

L song

nominal talker

it is easy to

Take,

writer.

A

conceive how

for example,

the agent

talker is a person who is

characterized by the act of talking (a lot,

usually);

writer

by

is a

person

who is

characterized

writing songs (professionally,

usually).

writing songs constitute an act, denoted by V', V'.

song

the act

of

Both talking and

which is what is typically

V is a possible expansion of

and after all,

So we can say that nominalizaton and adjectivization in

English with the suffixes -ing, -er, and Note

that

these

transparent

suffixes

among

the

are

the

various

-)5

can apply to V'.

most

productive

and

to

nominalize

and

nominalization

and

ways

adjectivize verbs in English. The

claim

that

adjectivization uncontroversial

certain

processes one.

On

fundamental assumption

English apply

to

V'

the contrary,

of the

compounds.

in

all This is

of above-mentioned one

not

an

it violates

the

lexicalist approach

phrasal category can enter word formation. maintained

is

of the

This position is

studies

reasons

that no

on

verbal

why the

simple

6 Kiparsky (1983), who basically supports Roeper and Siegel's analysis also collapses what in his formulation of rules correspond to (4a,b) and (5a,b) (not the Z-derivation of adjectives). However, in his lexicalist framework, he does not consider V' to enter his rules. He consequently has the same problem as the others in defining what X can be in [X V].

68

generalization that verbal

compound

had to

discussed above. the

In

consequence of

formation formation mentioned

a well-formed be

V' (V

captured

a

can

in other

Chapter I we have allowing

X)

derive a ways,

as

already touched on

phrasal category

in

word

In our rule typology scheme the verbal compound involves the there,

interaction

that

seems

characterization of the process.

of A to

and

give

C, the

and

as

correct

Namely, in the analysis by

Roeper and Siegel verbal compound formation seemed to have a dual characteristic of being syntactic (structure dependency in

its

condition)

and

of

being

lexical

(category-changing) according to Wasow's (1977) criteria. By viewing

it as

typology,

a

we can

process of say

A and

that the

C

interaction in

syntactic

nature of

our the

process lies in its phrasal input {A) and the lexical nature of the process lies in (C).

Indeed,

the

its being a morphological derivation productivity

and

the

semantic

compositionality of verbal compound formation are attributed not only to these particular

suffixation processes but also

to the phrasal expansion under V'. in (4)

Note also that our rules

are simple derivational rules and do not involve the

mechanism of transformation.

We will

explain below that a

morphological well-formedness

condition restricts

the form

of the output (e.g., number and order of elements). As mentioned earlier (Chapter I: English

a class

of complex

3.3.3.) there exists in

adjectives

which has

similar

69 properties

to

the

verbal compounds,

warm-hearted,

three-legged,

discussed the

type of problems

as

and so on.

exemplified We

have already

these compounds

have given

for the framework of level-ordering morphology (cf. I:

2.5.).

to

this

by

Chapter

Clearly, we can give a straightforward analysis class

of

compounds

in

a

similar

way

to

the

derivation of verbal compounds. (6) N'(X N)

+ ed ---> [[X]

[N-ed]] A

As in the case of verbal compounds, N (ex.

wing-ed),

meaning is:

and in

N' can also be a simple

all cases the basic compositional

characterized by the possession or the presence

of N'. Bresnan (1982a)

gives

a nice piece of

evidence showing

that the input N' in (6) must be a syntactically well-formed N'.

An

adjective which

position (ex.

cannot appear

in question:

this fact Bresnan suggests that condition on

concludes

that

*asleep-footed.

the morphological

"it is

incorporation of

single words," but then

unlikely

that

condition is a formal condition of grammar,

this

'global'

but rather must

be a factor that affects perception and lexicalization." cannot

accept

these

remarks

From

"syntactic adjacency may be

independent lexical categories into she

prenominal

my foot is asleep I *asleep foot) also cannot

appear in the compound

a general

in the

on

the

following

We

grounds.

First, syntactic adjacency is not a general condition of the process

of this

type,

as

evidenced by

verbal

compound

70 formation; elements

in the under

case of verbal compounds,

V'

must

be reversed

constraint of compounds (see below). adjective

formation

lexicalization of

is

a phrase

regard

this

(i.e.,

supplied

the

formal

Secondly, this complex different

lexical

from

unit formation

in that what is involved is Consequently,

and productive phrase.

example

fit

fundamently

belonging to C in our typology) a compositional

to

the order of

by

Bresnan

as

we

supporting

evidence for our rule (6). The

similarity between

complex adjectives in

the verbal

terms of their transparency

pointed out in Chapter I (cf. (11)). internally

take a

compounds and

comparative or

these

was also

The fact that they can superlative ending

warmer-hearted, earliest riser), for instance,

(ex.

supports our

proposal that these forms are derived by syntactic processes as

opposed

to

the

primary

compounds.

We

comparative affixation as inflectional (class C) often interact with phrasal have

the

comparative

affixation takes place; V'.

consider and, thus,

operations.' The examples above

marker

on

X'

level

[warmer heart] N',

before

the

[rise earliest]

For the complex adjectives, there naturally is another

slot for the comparative marker, entire compound:

with its domain being the

more warm-hearted.

'This applies also to the periphrastic markers more, most, with which the comparative endings alternate. ----- -----

71

We

have

thus

recognized

three

types

of

suffixation

processes that can apply to V' and one that can apply toN'. Although such

a claim

cannot be

consistent with

a strict

lexicalist position, postulating them does by no means allow unrestricted interaction between syntax and the lexicon. observed

in the

morphological properties,

discussion of operations

and

interaction

rule typology

typically

in many languages

between

derivational processes

exhibit

(such as

is quite limited.

in English,

postulated here

these

seem to

scheme, lexical

English)

(syntactic-phrasal)

A

(syntactic-morphological}

processes we

our

As

the

and

c

Among all the four suffixation

mark the

exception.

Further, even though they include V' and N' category,

there

seem

word,

to be

some

making them English.

formal constraints

parallel in form First,

affix must consequence,

the output

to the ordinary

they must be right-headed,

attach to

the head

V and

the order of elements

the opposite (X

on

V}

night

the modifier

I*

night

to the

head N

N.

As

a

in a verbal compound is

from that in V' (V

truck driver.

and hence the

the head

X).

compound cannot consist of more than two words: (at}

compounds in

8

must be

Secondly the drive truck

In complex adjectives, a single

word (or

a

compound): deep-blue eyed I* round (and) blue eyed.

8 0n the other hand, since compound formation is recursive, compounds themselves can be compounded with other words: cf. [coffee maker] [maker]. If this example above is interpretable, it is structured: [night] [truck driver].

72 1.4. No subject condition It has been mentioned that

people have observed what can

be called a 'no subject condition' on verbal compounds: in a verbal compound relation

the first element

to the

Selkirk {1982)

verb.

It

cannot bear

was pointed

had to state

the subject

out earlier

this condition as

that

a separate

observation because her observations on verbal compounds are based on their functional argument structure. achieved

a

parallel

effect

by

linking to internal arguments. this

condition

formation.

a

is significant

absolutely therefore

But

argument

for neither of them is of

the

compound

In our analysis this condition follows directly

from the rule because V' point

restricting the

consequence

natural

Lieber {1983)

This

cannot include the subject.

in that

exceptionless

this condition

for the

it distinguishes

verbal

verbal

seems to

compounds,

compounds from

be and

primary

compounds with similar meanings, as shown in the following. {7) a. b. c. d.

9

*heart failing I heart failure *population growing I population growth *rain falling I rainfall *earth quaking I earthquake

Note that Selkirk's notion of the items in {7),

for which

verbal compounds includes all her observation of 'no subject

condition' does not uniformally hold.

Similarly,

although

9 Note that only -ing nominals have the potential of incorporating the subject. In agent nominals the subject relation is expressed(?) by the suffix, and in passive compounds the subject relation is always outside the compound {typically the element modified by the compound): hand-made objects.

73 Kiparsky (1983) rules,

captures the

the contrast

admits),

for

he

contrast of

in (7c,d)

includes

is not

(7a,b)

by

his

explained (as

~-formation

verbs

in

he

verbal

compounds.

a

From semantic considerations,

it can be pointed out that

verbal compound

V'-ing

generic name

of

the form

for an act,

since

denotes.

On the other hand,

denote is

not an act,

but

usually denotes

that is what

what

a

V' generally

a subject NP and a verb

is an event.

It consequently

seems natural that even with primary compounds this relation is not commonly

realized (i.e.,

named),

and

many of such

examples seem to refer to some sort of natural phenomena, as exemplified in (7)

above.

with Japanese compounds Nevertheless,

2.)

This

tendency is also attested

of the similar kind.

the fact that verbal

express such cases is significant

(See section

compounds cannot

and supports our analysis

based on V'. 1.5. The generic condition on the realization of the argument structure We will how

the

now examine one interesting argument structure

lexically

states that compound spirit

observation

verb is

(b)

Lieber's

about

verbal

must be

This generalization (1983) ALP,

realized

Selkirk's

all non-subject arguments

structure. to

a

syntactically.

and

above-mentioned

of

condition regarding

which

both (1982)

compounds realized in

is

similar

states

that

in all

74 internal

arguments must

be linked

in compound

structure, Selkirk

although it applies in a very different framework.

(1982: fn.l5) admits the inadequacy of her generalization in the face of the examples supplied by Roeper, the

agent

appearing

argument

this example by

appear outside subject

saying that agent

the compound since

relation to

the verb

as

compounds it outside:

is possible for

hand-made

One

interesting

realization of

gy

arguments can bear the

cannot enter

the

case of passive compounds

Note that

teacher-trained.

compound:

they typically

and hence

compound structure except for the such

the

It seems possible, however,

cigarette-smoking Qy children. to handle

outside

which involves

even

with

the agent argument

passive to appear

children. fact

we

can

the agent argument

observe

this

about

with a by-phrase

is the

following contrast: (8) a. b. c. d.

Cigarette-smoking by children is upsetting. ?Cigarette-smoking by some students upset me. *Cigarette-smoking by my son upset me. *Cigarette-smoking by John upset me.

(9) a. Church-going by young people is on the decline. b. *Church-going by John pleases his mother. It

seems that

agent phrase

appearing

compound is largely restricted to

generics.

on the

the

other hand,

that the

with a

verbal

It turns out,

genitive realization

agent is not restricted in the same way: (10) a. John's cigarette-smoking upset Mary. b. John's church-going pleases his mother.

of the

75 It seems

that this

is due

marker can express case it

seems to

to the

fact that

the genitive

a wide range of relations,

and in this

express something

like 'John's

habit of

V'-ing.' It has

long been

nouns in a

noted (Postal

1969 and

others)

compound cannot be referential.

that

For instance,

they cannot be an antecedent to an anaphora: (11) *John is a horse thief, but at least he fed it/them. Similarly, Levi (1978) points out that a woman-hating editor "could only describe

an editor who hated

but not (atypically)

one particular

that what can be called extends to the

consequences this of

how

the

woman."

realized

agent argument of the

It is unclear to me at observation has for the

syntactic realization Nevertheless,

compound formation.

It seems then

a 'generic condition' for compounds

~yntactically

compound as well.

women in general,

should the

this point what general question be

related

to

generic condition

seems to be a salient generalization, and we find below that it extends to some other cases as well. There

are

a

couple

of

fairly

1 "

productive

adjectivize verbs, namely the suffixation of -able, un-passive

formation.

We

can

observe

that the

ways

to

and the generic

10 Roeper (1984, ms) discusses the syntactic and lexical realizations of thematic roles in GB framework. He talks about the appearance of the agent phrases with various deverbal constructions, but does not consider any particular constraints on them. His examples, however, gave me some hints for the observation made here.

76 condition

seems

to

be

applying

to

them

as

well,

as

exemplified below: (12) a. This game is playable by four people. b. *This game is playable by John.

(13) a. This problem is solvable by a ten-year-old. b. *This problem is solvable by John. (cf. This ploblem can be solved by John.)

(14) a. This phenomenon was undescribed by scientists. b. ?*This phenomenon was undescribed by Chomsky. (cf. This phenomenon was not described by Chomsky.) The

existence of

follow

from the

such

nature

a of

generic constraint

seems

these derivational

to

processes.

Verbal compounds, as mentioned before, express generic names of acts

(as opposed

-able or (as

un-passive adjectives

opposed to

semantically agent.

to specific

instances of express general

specific states),

congruous with

a

them),

and thus

specific

properties

they are

or a

and

not

referential

The generic nature

of these derivational processes

stands in contrast with the

more syntactic and periphrastic

means

of

expression:

verbal

compounds

vs.

gerunds,

un-passive vs. not + passive, -able vs. can be V-ed. We find another instance where be

considered relevant

in the

the generic condition may lexical

unit formation

complex adjectives discussed in Chapter I (3.3.1), they

are not

section.

related to

We observed that

the processes

their

infinitival

although

discussed in

this

easy-type adjectives and object

complement deletion adjectives can with

of

form a complex adjective

complements:

easy-to-please,

77 pretty-to-look-at,

ready-to-serve.

On the other hand,

was

equi-adjectives

cannot

noted

that

adjectives

in

the

*ready-to-walk,

same

way:

*willing-to-talk.

form

Among

other things one

between these two classes

is

the

nature of

complement. of

subject

complex

*eager-to-succeed,

salient difference the

it

PRO of

of adjectives

their

infinitival

Namely, the former class of adjectives have PRO

arbitrary

reference,

while

the

equi-adjectives

have

subject-controlled PRO: {16) a. John is easy

PROarb to please tl

b. Johni is eager

PROi to succeed ]

If we can assume that PROarb a controlled

correspond to a generic NP and

PRO to a referential

NP,

we can

explain the

facts here as another instance of the generic condition. 2. Deverbal nominals and compounds in Japanese The

basic

nominal

form

of

a verb is the

infinitive form {renyoo-kei) in Japanese: when the root ends with

~.~,no

form participates in various as

a nominal

infinitive. two forms

as well

as

As we will see result in

same as the

verb root + i, or

overt suffix is added.

This

productive compound formations in syntactic below,

cases where

environment as

an

the homophony of these

it is

hard to

determine

whether a certain occurrence of this form is a nominal or an infinitive.

The

following list shows

the basic

types of

78 derivations in which this form of a verb appears.

11

(17) simple nominalization: VN (deverbal nominal) a. asob-u --> asob-i 'play' b. (applies to a complex verb form of V-inf. + V) uke-ire-ru (receive-take in) --> uke-ire 'acceptance' (18) compounding of VN's: [VN VN] a. kiri-uri (cut-sell) 'selling by pieces' (cf. hakari-Ori (weigh-sell) 'selling by the weight' b. awate-gai (hurry-buy) 'hasty purchase' c. hasiri-gaki (run-write) 'scribble' (19) dvandva compounds of VN's a. uri-kai (sell-buy) 'selling and buying, trade' b. yuki-ki (go-come) 'going and coming, traffic' (20) [X VN] compounds a. [N VN] yuki-doke (snow-melt) 'thawing of the snow' b. [N VN] kutu-migaki (shoe-polish) 'shoe polish(er)' c. [ADV VN] haya-oki (early-rise) 'early rising/riser' (21) [VN X] compounds a. hosi-gusa (dry-grass) 'hay' b. kai-mono (buy-thing) 'shopping' (22) reduplication a. [VN VN]A tiri-ziri (scatter) 'all scattered' b. [V-inf. V-inf.]ADV kaki-kaki 'while writing' (23) suffixation a. (adjectival) aruki-tai (walk-want) 'want to walk' b. (nominal) aruki-kata (walk-way) 'way of walking'

11 There is a phonological process called 'rendaku' (literally 'sequential voicing'), which changes the voicing of the initial voiceless consonant of the second element under certain conditions. In (18b), for instance, kai is changed to gai in the compound structure. For--some observations on this process in relation to compound formation, see Section 3.

79 In what

follows we

deverbal nominal Section 2.2.

will first

consider the

compounds (20)-(21}

we will look

properties of

in Section

2.1.

In

at nominal suffixation (23}

and

some related constructions. 2.l.Deverbal compounds

2.1.1. ~ of deverbal compounds Compound formation has generative grammarians

been given some attention

working on Japanese.

sketches a transformational analysis has a sentential

source.

Okutu (1975}

in which each compound

Makino (1976}

syntactic approach to compounds because of semantic idiosyncracy involved

by the

argues

against a

of the large amount

in compounds.

The topic

is taken up again recently in Kageyama (1981), where he also argues things,

against transformational he points

out that the verbal

calls 'nominalization

compounds')

wider range of meaning than and

also

noun'

illustration,

let us

can

Among

other

compounds (which he

in Japanese have

a much

their English counterparts with

that the

'incorporated

compounds,

derivation.

g·rammatical be

much

more

relations varied.

look at the possible range

classifying them by the

of

the

As

an

of [X VN]

meaning of the compound

and the function of the first element: (24) types of (X VN) compounds ( +: examples from Kageyama A. Agent a. DO: mono-tori (thing-rob) 'thief' kane-moti (money-own) 'wealthy person'

1981 )

80

B. Instrument + a. DO: nezi-maw~si (screw-turn) 'screw driver' kan-kiri (can-cut) 'can opener' C. Result/Product + a. DO: yasai-itame(vegetable-f~y) 'fried vegetable' b. SU (Vt): musi-kui (bug-eat) 'a hole in cloth, etc. caused + by a bug eating it.' c. SU (Vi): mizu-tamari (water-accumulate) 'puddle' d. by-Agent: musi-sasare (fug-be stung) 'bug-sting' e. Instrument: kami-zutumi (paper-wrap) 'something wrapped with paper' D. Place/time a. DO: mono-hosi (thing-dry) 'a veranda for hanging laundry' b. SU (Vi) : hi-gure+ (sun-set) 'sunset (time) ' E. Act(-suru)/ Eve~t(-ni naru,'become' -ga aru'there is-' a. DO: kane-mooke (money-profit) 'making profit' b. SU (Vi): yama-kuzure (mountain-collapse) 'avalanche' c. SU (Vt): kami-kakusi (god-hide) 'mysterious dissapearance' d. IO: hi to-makase+ (person-depend on) 'being dependent' e. Goal: tera-mairi (temple-visit) 'temple-going' f. Source: ie-de (house-leave) 'running away from home' g. Instrument: suna-asob~ (sand-play) 'playing with sand' h. Time: yo-asob~ (night-play) 'go out and have fun at night' i. Adverb: waka-zini+ (young-die) 'early death' j. 'like aN': kaeru-tobi (frog-jump) 'jump like a frog' F. Nominal predicate (-da 'be X') a. DO: oya-nakase (parent-cause to cry) 'being a bad child' b. Goal: gaikoku-yuki ~foreign land-go) 'foreign bound' c. Source: huransu-gafri (France-return) 'returnee from F.' d. Time: Meiji-umare (Meiji era-born) 'Meiji generation' e. Place: Osaka-sodati (Osaka-grow up) 'have grown up in 0.' Although this vast range When

we

is not an exhaustive

of meanings expressed consider

the

fact

list,

one can

by the [X that

they

see the

VN] compounds. are

formally

non-distinct from each other, it is not surprising that some of them

can be ambiguous depending

for instance,

on the context

kutu-migaki (shoe-polish)

of use:

can be Agent 'shoe

polisher', Instrument 'shoe polish' or Act 'shoe polishing'. Note that the English counterparts are marked with different

81

suffixes according to their meanings. Compared with less variation.

[X VN] compounds, They can be

12

[VN X]

compounds show

classified in two categories:

Act nominal and modified nominal. (25) types of [VN X] compounds A. Act nominal DO: wasure-mono (forget-thing) 'forget/leave something' oki-tegami (leave-letter) 'leaving a letter (for someone to read)' yaki-mono (bake-thing) 'pottery' B. Modified nominal (N that (is) V-ed) hosi-gusa (dry-grass) 'hay' yaki-zakana (bake-fish) 'broiled fish' nokori-yuki (remain-snow) 'remaining snow' asobi-ba (play-place) 'playing ground' kiraware-mono (be disliked-person) 'unpopular person' Although (25A)

and

(25B)

are formally similar,

former can be used in the frame 'VN-o suru', are all concrete nouns.

only the

and the latter

For both of them, only N can be the

second element, unlike [X VN] compounds. 2.1.2.Analysis for the deverbal compounds We can compounds simple

observe from

(24)

involving deverbal

analysis

parallel

deverbal compounds.

to

and

(25)

nominals what

we

that cannot gave

the Japanese be given for

a

English

It was claimed in Section 1 above that

English deverbal compounds are derived by suffixation to V',

12 English nominalizing suffixes do carry multiple meanings: ex. -er: Agent/Inst., -ing: Act/Product, etc. Nevertheless, compared with the Japanese cases, they contribute in distinguishing the possible meaning of the deverbal compound. See below for more discussion on the function of suffixes.

82

and

that

compound.

FSP restricts

items

that

can form

a

deverbal

Japanese compounds in (24) and (25), on the other

hand, do not appear to have such restrictions.

We can spot

a number of cases which would violate FSP:

the non-deverbal

element can be

and intransitive

subject of both transitive

verbs (24E.b,c),

Indirect object

transitive verb (24C.e), Japanese has a

V',

and so

and if it

(24E.d), on.

Instrument of a

The issue of whether

does,

on what level,

is a

controversial one, but given such diversity of data here, it is clear that one cannot restrict the items in the compounds using such notions as V' or FSP. This is not to say that there is no pattern at all in the way these compounds are formed. and DO

are the most

different types (24B),

Actually, transitive verbs

frequent combinations,

Agent nominals (24A),

and [VN-N) Act nominals (25A)

as the non-deverbal element.

and

among the

Instrument nominals seem to allow only DO

There is also a certain amount

of correlation between the type of relationship V and X hold and the way the compound is unlikely type

of compound

used. [SU,

For instance, Vt) cannot

a rather

be used

as an

ordinary action nominal but rather is used in a passive-like context with the direct object of the verb as the subject of the sentence: (26) a. Kodomo ga kami-kakusi ni at-ta. child NOM god-hide DAT encounter-PAST 'A child disappeared (as if taken by God).' b. Kare wa hito-zuki ga suru. he TOP person-like NOM do 'He is likable.'

83 c. Kare wa hito-warai ni na-tta. he TOPperson-laugh at DAT become-PAST 'He became a laughing stock.' It is also interesting that the cases

of [SU Vi] seem to be

limited to some types of natural phenomenon: (27) ame-huri (rain-fall) 'rainfall' zi-suberi (land-slide) 'landslide' kaza-muki (wind-turn) 'the direction of the wind' hi-deri (sun-scorch) 'drought' These

compounds correspond

$-suffix

that

was

to the

discussed in

English compounds relation

to

the

with 'no-SU

conditon' in Section 1, such as sunset, landslide, snowfall, and so on.

13

We are not going to discuss the semantics of each type of compound, but this brief discussion of some of them leads us to think

that it is

semantic

category of

English

the lack of the

a suffix to

compounds

deverbal compounds)

indicate the

(unlike the

that seems

to allow

case

of

various

usages and hence the wide range of the relationships between

v

and X.

mismatch

This lack of suffix also is between the

category

element and the compound itself. Kageyama (1982),

of

responsible for the

the morphological

head

In regard to this point,

after stating that the compounds should be

derived in the lexicon, assumes that a lexical reanalysis of the following sort often takes place.

13 Mikami (1953) makes the remark that the subject NP and a passivizable verb ('noo-doosi' Vt and some Vi's) cannot form a compound, and only non-passivizable verbs ('sho-doosi' some Vi's) can form a compound with the subject NP.

84 (28) a. [kane-moti] VN-->N (money-own) 'wealthy person' b. [kai-mono] N-->VN (buy-thing) 'shopping' Although this categories,

account can adjust

the frequent

mismatch in

it allows arbitrary switching between N and VN,

which is not plausible in view of the observations below. Looking again

at examples of

various compounds

in (24)

and (25), we see that the cases where this mismatch does not occur are [X VN] action nominals VN (-o suru), used

as

N.

and [VN N] Now,

(24E)

that can be used as

modified nominals (25B)

except for

those

that

idiosyncratic usages (as exemplified above), VN] compounds understood

can be

Z-head.

1

that are

have

rather

the rest of [X

straightforwardly seen

as having

an

~

(29) A, Agent: kane-moti-~ B. Inst.: tume-kiri-~ c. Result: musi-kui-~ D. Place: mono-hosi-~

(money-own) 'wealthy person' (nail-cut) 'nail-clipper' (bug-eat) 'a hole made by bugs' (thing-dry) 'a veranda for drying laundry' E. Time: hi-gure-% (sun-set) 'sunset time' F. Act: ie-de-~ (home-leave) 'running away from home'

In fact the

transformations proposed by Okutsu

Makino (1976) approach)

(1975)

and

(although the latter eventually rejected that

derive these

compounds from

a relative

clause

structure with an abstract head: i.e., [kane o motu] hito 'a person who

has money'.

By postulating

this if-head

of N

14 Although [X VN] act nominals are mostly used as VN's, they are also used as N, as shown below: (i) kane-mooke o suru I (*kane o mookeru koto o suru) 'to do profit-making' 'to do the act of profit-making' (ii) kane-mooke wa tanosii./ (kane o mookeru koto wa tanosii) 'profit-making is fun the act of profit-making is fun (i) shows the usage as VN, (ii) as N.

85 category which

and leaving

the rest

applies according

lexically

listed for

to

each

to

the

output semantic

item (the

lexically listed in any case), deriving them

an interpretation

category

rule

category has to

be

we can avoid the problems of

from the relative

clause source

(see Makino

1976) and still maintain the generalizations.

(30)

[kane-moti]VN~

This analysis

]N,Agent --> 'kane o motu hito'

is supported by

the fact that

sometimes the

head is overtly realized as exemplified below. (31) a. tume-kiri-basami (cf.29B) (nail-cut-scissors) 'scissors for nails' b. mono-hosi-ba (29D) (thing-dry-place) 'a place for drying laundry' c. hi-gure-doki (29E) (sun-set-time) 'senset time' Now that

we have claimed that

all the [X

VN] compounds

are headed, we are left with the [VN N]VN compounds (25A) as the only case of the category mismatch.

There is actually a

well-motivated reason that this should be so. compounds are compounds

formed by analogy

which preserve

language, Chinese: ex. the Since

native compound SJ

compounds

the

to the [V 0]

Namely, these

Sino-Japanese {SJ)

order

of the

source

satu-zin (kill-person) 'murder' {cf. hito-gorosi (person-kill) occupy

an

important

'murder').

part

in

the

vocabulary of Japanese it is

not surprising that their word

order influenced some of the

native compounds.

15

The native

15 Nishio (1976) gives the interesting observation that the V-0 order of the SJ compounds cannot extend to the compounding of SJ compounds themselves, which always has the 0-V order: satu-zin (kill-person) 'murder'

86 compounds

of

the

form

[VN

N]VN

are

limited

to

the

combination of V and DO, and even within that limit not very productive,

this limited class of compounds, then, a feature [+SJ], head.

It is only for

compared with the [N VN] kind.

that we need to have

which assigns the left element

This feature is needed

of genuine SJ compounds.

to be the

anyway for the interpretation

Consequently,

we do not need to

assume an unmotivated lexical reanalysis rule (28) by Kageyama (1982).

All Japanese compounds are right-headed

except for the ones with the be a

proposed

[+SJ] feature.

natural result considering

all phrases are right-headed,

the fact that

This seems to in Japanese

because all arguments precede

Vandall modifiers precede the heads.

16

boo-si (prevent-stop) 'prevention' satuzin-boosi 'prevention of murder' *boosi-satuzin This is due to the fact that each morpheme that forms a SJ compound (satu, zin, etc.) is largely bound, but the compound itself (satu-zin) is a free word. So the compounding of compounds is more transparent, and, unlike compounding of morphemes, can have a phrasal counterpart: satuzin no boosi, which naturally takes the 0-V order. Incidentally, the compounding of SJ compounds is by far the most unrestricted form of WF process in Japanese (due to the lack of morphological ending and restrictions in combination). Okutsu (1975) gives an example of a SJ compound with 26 chinese morphemes. 16 Bloomfield (1933: 14.5) discusses the classification of compounds with the terms 'syntactic' and 'asyntactic'; in syntactic compounds members 'stand to each other in the same grammatical relation as words in a phrase', while the members of asyntactic compounds 'stand to each other in a construction that is not paralleled in the style of their language'. Thus we can say for Japanese compounds with

87

Concluding the observations so far, basic

difference

Japanese VN

between English

compounds is the

we can say that one

deverbal

compounds

lack of explicit

and

suffixes in

the latter which makes the semantics and usages of them less predictable by means of derivational rules. 2.1.3. Abstract nominals in VN compounds .-- -Another

noticeable difference

compounds and

their Japanese

occurrence of

abstract nominals

derived nemur-i

from

intransitive

'sleep',

aruk-i

'laugh',

odor-i

'dance',

nominals

derived

from

between English

counterparts is in the

verbs

'walk',

the frequent

latter.

are

Nominals

straightforward:

oyog-i 'swim',

and so on.

transitive

deverbal

On verbs

wara-i

the other hand, are

much

less

frequently used independently. 17 To

illlustrate this point,

let .us

in

consider the

noun

writer

English,

as

in

a

sentence:

deverbal nominals that they are mostly syntactic except for the ones with the [+SJ] feature. 17 Besides appearing in compounds of the type discussed in the present section, a nominal derived from transitive verbs can combine with other deverbal nominals and form different types of compounds (cf. (18} (19}}, in which case DO can be expressed with a genitive NP: hon no uri-kai 'selling and buying of books' (cf. *hon no kai, *hon no uri}. It is suggested in Nishio (1961} that the single deverbal nominal in general is too 'unstable' in form and meaning to be independently used, because there is no suffix to indicate its meaning, and it is phonologically short (usually one or two syllables}. This speculation, although it seems to be partially correct, does not explain the fact that the nominals derived from intransitive verbs are much more frequently used by themselves than those of transitive verbs.

88 (31) John's father is a famous writer. Here DO of writing is unspecified and does not appear. fact is pointed out and captured in Levi (1978: rule of unspecified NP non-head NP's: argument x expressed

deletion,

as a

PP

[of x].

5.3) by the

which deletes unspecified

in the case of writer,

in WRITER

This

it deletes the theme

The theme

('writer of

this

deverbal compound ('song writer'),

argument may

article')

be

or in

a

as we have already seen.

The following are examples of VN compounds in Japanese where the DO is incorporated: (32) uta-kaki 'song writer', syoosetu-kaki 'novelist' shibai-kaki 'playwright', kyakuhon-kaki 'script writer' komaasyaru-kaki 'commercial writer' Now,

unlike

English,

corresponds to rather, 'writer'. that

~here

is no

an agent nominal

the abstract

Japanese that

with null

the corresponding form is: This form

form in

theme argument,

mono-kaki (thing-write)

is parallel to those

noun mono

is

in (32)

there to

fill the

except theme

argument slot without contributing any substantial meaning. In the following according to the

some examples of mono-

VN's are given,

position of the unspecified

case frame of the verb. (33) A. Accusative Object (x-ga y-o V) a. mono-goi (thing-beg) 'begger, begging' b. mono-uri (thing-sell) 'salesman' c. mono-tori (thing-rob) 'robber, robbery' d. mono-siri (thing-know) 'learned person' e. mono-moti (thing-own) 'wealthy person' f. mono-oboe (thing-memorize) 'memory' g. mono-mane (thing-mimic) 'mimicry' h. mono-osimi (thing-spare) 'being stingy' i. mono-mi (thing-see) 'sight-seeing'

object in the

89

j. mono-omoi (thing-think)n 'meditation' B. Dative object (x-ga y-ni V) a. mono-ozi {thing-be shy) 'being shy' b. mono-nare (thing-be accustomed) 'being accustomed' c. mono-aki (thing-be bored) 'being bored'

c.

Nominative object (x-ni y-ga V) a. mono-iri (thing-need) 'being in need' b. mono-wakari {thing-understand) 'understanding' c. mono-zuki (thing-like) 'being curious {lit. like strange things)'

As we

can see from

this list,

these are all

accusative NP's being the most frequent. act nominals,

and

theme NP's,

Many of these are

they often combine with suru

'do' to be

used as a verb, in which case the verbal expression is quite similar

in

function

to

the

'intransitive

use'

of

a

transitive verb in English, as shown in the examples below: (35) kare wa mono-goi o site kurasi-ta. he TOP begging ACC do live-PAST 'He begged for a living.' We can see Japanese is

that one function of these to provide

deverbal nominals in

an 'intransitive'

form {where

the

object NP is unspecified) for a transitive verb. We noted above that there are

two types of act nominals,

[X VN] and [VN N], and that the latter is left-headed due to the analogy to SJ compounds. is indeed a

special case of deverbal

expect the abstract same way,

If the latter [VN N] compound

and that

nominal compound,

theme nominal to be is what we

find.

functioning in the See

the following

examples. (36) a. b. c. d.

we

kaki-mono {write-thing) 'writing' kai-mono {buy-thing) 'shopping' ami-mono {knit-thing) 'kntting' arai-mono {wash-thing) 'washing, laundry'

90

e. okuri-mono (give-thing} 'gift-giving, gift' Mono

in these

NP's.

examples

also

indicates unspecified

theme

The fact that nominals in (34) and those in (36) are

used on a par with each

other despite the difference in the

order of elements can be seen by the following passage taken from actual text, where they are conjoined by

'and'.

~

(37} Konogoro otosi-mono ya mono-wasure ga hagesiku, recently lose-thing and thing-forgetNOM frequent

sagasi-mono o site ite mo nani o sagasite iru no ka search-thingACCdoPROG even whatACC searchPROG COMP wasurete simau. forget completely 'Recently I frequently lose things and forget things, and even when I am looking for something, I forget what I was looking for. It

was mentioned

section

that the

structure

of the

difference between

a

number of

times

deverbal compounds verb.

It is

in the

reflect the

not

preceding argument

surprising that

English)[-writer and

the

Japanese mono-kaki

is also found in the sentential structure, as shown below: (38) a. He writes to earn a living.

b. Kare wa *(mono o) kaite kurasite iru. he TOP thingACC write live PROG Hence it is clear that the existence or non-existence of the rule on unspecified

object deletion should be

the argument structure of the verb itself,

specified in

rather than as a

part of the nominal formation rule as done in Levi (1978).

91 The

'intransitive' use

of

a

transitive verb

unspecified object is discussed and of

intransitivization

framework. argument

in

It assigns (ex.

write

{SU,

(l982a) function

~)),

an

accounted for by a rule

Bresnan the null

with

while

in

the

to the

LFG

object

not changing

the

predicate argument structure (write (Agent, Theme)), so that the null object argument is still existentially bound in the semantics. type of

This type

of rule is in

rules such as

predicate argument break (Theme).

contrast with another

inchoativization,

structure:

It can be

which

break (Agent,

alters the Theme)

-->

said that while inchoativization

changes the semantic valency of the verb, intransitivization leaves it

unchanged.

In Japanese,

there seems to

be no

productive application of this intransitivization rule, this

is

Namely,

reflected

in

the

form

of

deverbal

and

compounds.

the abstract nominal mono is used as a 'filler' for

the unspecified object argument both in sentential structure and

in

compound

intransitivization

structure, rule

is not

The

fact

completely

that

productive

the in

English is seen from the use of the corresponding word thing in English

in the following

type of sentences,

which are

taken from Fronek (1982). (39) a. She began to notice things. b. I am always forgetting things. It

seems then

transitivity

that is

this

not as

opposition of essential

as

(V. Woolf) formal the change

(surface) in

the

92 predicate

structure,

argument

It is

inchoativization rule. Japanese, the

represented

by

rather significant

the

that in

while the intransitivization rule does not exist, opposition

inchoative/transitive

marked by ('break,

as

the verb

morphology,

unlike

tr.')/ kowar-eru ('break,

tr.')/ yak-eru ('burn, that Japanese verb

intr.' ),

is

systematically

English:

intr.'),

etc.

16

kowas-u

yak-u ('burn,

We can observe here

morphology is sensitive to

and reflects

the change in the verb's predicate argument structure, while English shows little Another instance

of such contrast

derived from an John

walks

morphology on the verb

dog.

causative suffix:

The

aruk-u I

that the absence of the

the

nature of

'causative' verb

intransitive verb in English:

his

as reflected

is the

to reflect it.

has

a

We may speculate

aruk-ase-ru.

of deverbal compounds

Japanese verb of a verb,

changes the 'surface'

counterpart

intransitivization rule in Japanese

in the form

shows the valency

Japanese

John walks/

stems from

morphology

that it

and hence,

a

directly

rule that only

transitivity of the ve-rb

such as the

intransitivization rule does not exist for Japanese.

18 See Jacobsen (1980) for a detailed discussion of this morphological opposition.

account

and

93

2.2. Suffixed deverbal nominals

2.2.1. Nominalization suffixes In

this section

we

will discuss

a

few suffixes

that

attach to the infinitive form of a verb to form a noun. (cf. 23b)

One such suffix -te (a bound morpheme)

makes an agent

nominal, as exemplified below: (40) a. b. c. d.

oyogi-te 'swimmer' utai-te 'singer' kaki-te 'writer' (cf. mono-kaki) uri-te 'seller' (cf. mono-uri)

This formation of agent nominals differs in a number of ways from the in

compound formation of agent

40c,d)

First,

that

this

were discussed process

is

in

nominals (exemplified the previous

productive

transparent, unlike the compounds.

and

section.

semantically

Secondly, it can attach

to activity verbs regardless of their transitivity. -te agent

nominal is

formed from

a transitive

When a

verb,

the

object NP is either explicitly expressed by the genitive NP, or it has

to be anaphorically understood

while the

compound agent

nouns incorporated.

nominals cannot

in the discourse, have referential

The following examples illustrate this

point. (41) a. Karewa mono-kaki I *kaki-te da. 'He is a writer.' he TOP writer writer COP b. Kare ga kono geki no *mono-kaki I kaki-te da. he NOM this play GEN writer writer COP 'He is the writer of this play.'

94

c. Dare ga kana geki o kaita? 'Who wrote this play?' whoNOM this play ACC wrote Kare ga *mono-kaki I kaki-te da. 'He is the writer. he NOM writer writer COP In short,

denotes the agent role of

VN-~

when we use it we assume

former

a pair like is somebody

VN nominals denote a profession

+

agent in a habitual

said for

and

a certain activity that this agent

is involved in, while mono or an

that verb,

action.

The same thing

mono-uri 'salesman' and who makes

a living

can be

uri-te:

by selling

the

goods,

while the latter is somebody who has something for sale. In English there

is one respect in

nominals and )i-derived agent differ.

which suffixed agent

nominals have

been noted

to

The former systematically allows the realization of

the theme argument

by an of-phrase,

while

the latter does

not. (If it does, it is idiosyncratic.) (42) a. *the thief of the car b. the robber of the bank c. *the cook of the stew (cf. the cooking of the stew) This was pointed out by

Roeper (1983),

where he attributes

this difference to the presence of the affix, which, system,

preserves the 'thematic grid' (i.e.,

the thematic roles) supposed

to have

of the verb. this property.

speculates that it is the as

~.

-ing, roles

nominal,

and thereby

assignment of suffix is not

Roeper (1983)

further

presence of certain suffixes such

that enables

thematic

The null

in his

associated

the language learner to induce with the

source

verb

predict such differences as

of

the

shown in

95

( 42).

Although this

speculation

tested on a wider range of that we can find the

data,

has

to be

empirically

it is interesting to note

same difference between suffixed agent

nominals and compound agent nominals (as well as SJ compound agent nominals): (43)

a. *sinario no mono-kaki 'writer of scenario' b. *sinario no sakka (SJ) c. sinario no kaki-te

Note that while English theme argument realized in

the compound structure,

in (42)

can also be

in Japanese

only the

compound nominal allows it: (44} a. car thief b. bank robber c. sinario-kaki 'scenario writer' d. sinario-sakka (SJ) e. *sinario-kaki-te The fact shown by (44e) seems to hold even for a lexicalized -te

agent

professional

So,

nominal.

utai-te

singer (contrary

to the

generally

means

general rule),

a but

cannot make a compound in contrast with a SJ compound of the same

meaning,

unacceptability inappropriateness

of

(44e) of the

which

indicates

is

due

not

second element

to

that

the

the

semantic

for the

compound

structure. (45) a. zyazu no utai-te I *zyazu no kashu 'singer of jazz' b. *zyazu utai-te I zyazu-kashu 'jazz singer' We will later discuss the difference between compounding and suffixation in some detail.

96

Going

back

to

the

question

thematic role assignment, (42)

above

if

modification

we

of

the

transparency

we can account for

relate

structure.

this

suffixation

Namely,

the

in

the facts in to

a

noun

following

are

equivalent: (46) a. sinario no kaki-te 'writer of a scenario' b. [sinario o kaku] hito 'person who writes scenarios' We can

view (46a)

and (46b)

as consisting

elements with the difference in the head;

one with a

of the

same

the morphological nature of

full noun,

the other with

a bound

suffix. (47) a. [sinario kak-]-te b. [sinario kak-] hito In (48a) below, inflected as a

since the head is a full noun, finite form and the

is also assigned.

In (48b)

the verb is

accusative case marking

the bound morpheme attaches to

the VN (infinitive form) and accordingly the object argument takes a genitive marker,

which is the case any NP must take

in order to modify a noun. 19 (48) a. [sinario o kaku] hito b. [sinario no]NP [kaki-te]N

19 If it is a grammatical argument (marked by~, Q, n1 1n a phrase), the NP simply takes the genitive marker no,--but if it is an oblique argument such as kara 'from', ~ 'to', made 'until', and so on, the genitive marker attaches to the postposition: Tookyo kara no kyaku 'guest from Tokyo' (cf. *Tookyo kara kyaku).

97 According to our rule typology, be syntactic,

while (4Bab)

the formation of (48)

differ

in that (48a)

must

involves

only a phrasal operation and (48b) involves both phrasal and morphological

operations.

This

analysis

captures

the

parallel compositionality between (48a) and (48b) and at the same time the difference in their forms. One

might argue

that by

assuming that

preserves the thematic role assignment capture the same parallelism,

and

the suffix

of the verb,

-te

we can

then we can contend that

the suffixed agent nominals and compound agent nominals both belong to the lexicon except that they differ on this point. There are move.

several reasons

First,

there

productivity of

evidence,

the suffixed agent

given

'productivity'.

is not

a satisfactory

is the semantic transparency

the compound nominals, of

why this

although it the lack

nominals as

of clarity

of notions

45)

suggests that

type of derivation from the Our assumption

ft-nominalization and the compound formation. suffixation

g-nominalization

and

rules is consistent

is

a

syntactic

the compound

V'.

while are

(See

the

lexical below for

Thirdly, the most revealing

evidence is the occurrence

certain type of

rule

formation

with this observation.

more discussion on this point.) piece of

like

Secondly, the fact that this nominal cannot

this suffixation is a different

this

opposed to

is not a decisive piece

enter a compound as mentioned above (44,

that

and the

of this suffix

The type of phrase

with a

that appears with

98

the

agent nominal

where

x

is

a

is [x-ni

suffix -te

subcategorized

complement

of

Although this verb 'naru' appears with two NP's, complement

x,

the verb

forms

a

x' ,

nar-] 'become the

verb.

SU and the

tighter unit

with

the

complement than with SU. 20 This makes sense when we consider that the

complement is

not a

grammatical argument

of the

verb in the way subject and object are. Thus 'naru' is not a transitive verb even though it

occurs with two NP's.

When

the suffix -te attaches to this verb, the complement remains unchanged, as shown in the following examples. (49) a. Kare wa [yome ni nari]-te o sagasite-iru. he TOP brideCOPbecome ACC look for-PROG 'He is looking for a bride-to-be.'

b. Inaka dewa [isya ni nari]-te ga sukunai. countryLOC doctorCOPbecome NOM few 'Few become a doctor in a rural town.' Note

that

the bracketing

[yome

nil

possible configuration in Japanese, noun must

take the genitive

marker.

[nari-te] is

not

a

since an NP modifying a (See fn. 19)

addition, there are cases cited in Martin (1975:

In

224) which

involve a phrasal concatenation of verbs, 'V-te kureru' ('do a favor of V-ing'):

20 This point can be ·demonstrated by the fact that the complement NP resists scrambling to the S-initial position, while other grammatical arguments or adverbs do not: (i) Taroo wa isya ni nat-ta. 'Taroo became a doctor.' TOP doctorCOP become-PAST (ii) ??Isya ni Taroo wa nat-ta. The marker 'ni' on the complement is the adverbial form (ren-yoo-kei) of the copula.

99 (50) a. [Aite ni site kure]-te ga nakat-ta. companyCOPmake give NOM NEG-PAST 'Nobody would be my company.' b. [Moratte kure]-te ga arimasen. take give NOM be-NEG (polite) 'Nobody would take it.' In

(50a)

'naru'.

'suru'(site) These phrases

structures,

is

the

causative counterpart

never find their way

the domain

of

some nominal

The preceding observations

process

nominal as

suffixation

opposed

to the

This fact indicates

suffixes

certain cases beyond the lexical level.

agent

into compound

but with this suffix and several other ones (as

shown below), they are rather frequent. that

of

can expand

21

lead us to conclude is

a

in

that the

morphological-syntactic

compound

formation

which

is

morphological-lexical, and the former can interact with some phrasal-syntactic processes,

as illustrated

in the

chart

below according to our rule typology,

21 0ne conceivable alternative way of dealing with such cases as (49) and (50) is to claim that the complement is 'inherited' under the nominalization process. Moortgat (1983) attempts to provide the compositional semantics for the complement inheritance cases in English (e.g., John's willingness to please; John's kindness to Mary) by introducing a new function in the semantics which allows -ness to have scope over the infinitive or the PP while adhering to the lexicalist assumption that the suffix attaches only to the adjective in the word formation. We will discuss this approach in Chapter III in comparison with the phrasal suffixation approach that we adopt. It suffices to note here that while the English cases exemplified here do not build any new structure (they are [[NP] [PP)]NP), the Japanese cases discussed above do, due to the strict morphological condition that all the prenominal modifier NPs must take the genitive marker (see fn. 19.}

100

(51) lexical

syntactic phrasal: [SF o kakul hito [yome ni naru] hito

SF-kaki, SF-sakka

morph'l: [SF no] [kaki-te] [yome ni nari]-te The assumption syntactic

rules

compound of

VN's,

that the lexical rules is

consistent which are

with

[+N),

generally precede

the

fact

cannot feed

that

the

the agent

suffixation, since only the infinitive form of a verb can be suffixed. (52) a. yomi-kaki 'reading and writing' --> *yomi-kaki-te b. SF-kaki 'SF writing' --> *SF-kaki-te c. hasiri-gaki (run-write) 'scribble' --> *hasiri-gaki-te (cf. uke-toru v. 'receive' --> uke-tori-te) On the other hand, since a syntactically derived word can be lexicalized, nominal

to

it

is

acquire

'professional singer') (cf.

possible for a

special or to

the

are

several other

feed the

way as the agent

discussed here are concerned.

suffixation utai-te

compound formation

22

nominalization

attach to the infinitive form of in the same

(cf.

meaning

kai-te sizyoo 'buyer's market.)

There

agent

the verb,

suffix,

suffixes

that

and they behave

as far as

the facts

They are exemplified below,

along with their phrasal counterpart;

some of them can also

22 As noted earlier, some compound formation is more transparent and productive than others. One productive type is the act nominal formation by the incorporation of DO, and a compound such as the following seems to be acceptable: (i) sinario no kaki-te sagasi 'sinario-writer-to-be search'

101 be used as a free word as indicated in the parentheses. (53) -kata 'the manner of v-ing, how to V' ·a. hon no yomi-kata 'how to read books' bookGENread-way b. hon o yomu hoohoo (54) -kai (also a free word) 'the value of' a. kenkyuu no si-gai 'the worthiness of research' researchGENdo-value b. kenkyuu o suru kai (55) -sama 'the manner, appearance of' (also a free word) a. iki-zama 'the way a person lives' live b. ikiru sama (56) -sama 'on the point (in time) on V-ing' a. booru no suimen ni ukiagari-zama (o tataku) ball GEN surfaceLOC float-point {ACC hit) 'hit the ball when it comes up to the surface' b. booru ga suimen ni ukiagaru tokoro (o tataku) One suffix '-yoo' has an interesting distribution in that it leaves the accusative marker on DO unchanged when it is used as part

of the fixed

expression -yoo

~

nai 'there

is no

V-ing' as shown below: (57) a. kane no I *o tukai-yoo o kangaeru. moneyGEN ACC use-way ACC think 'think about how to spend money' b. kane no I o tukai-yoo ga nai moneyGEN ACC use-way NOM be-NEG 'there is no way of spending money' In (57b) when the accusative marker is also acceptable, expression

-yoo

~

nai

is very

likely

auxiliary with the meaning 'cannot'.

reanalyzed as

the an

102 These close

examples show

(in derivation

how a

and perception) in

modification

structure,

formation.

Classifying the

latter as lexical,

nominal

despite

contrast former as

suffixation can to

a phrasal

with

the

be noun

compound

syntactic and

the

their both being morphological,

can capture this difference nicely. 2.2.2. Deverbal nominals of ambiguous status We have just mentioned one with

a V'

(NP

V-inf.-yoo)

o

V-inf.-yoo)

as

well

which can occur as

aN'

(NP

when it is part of the fixed expression -yoo

There exists

nai.

suffix -yoo,

a more frequent and

no ~

productive type of

construction with a similar property, as exemplified below: (58) a. oyogi ni iku 'go swimming' b. asobi ni kuru 'come to play' c. kowasi ni kakaru 'set about to destroy' Morphologically,

the verb

ambiguous between an

form (of the first

infinitive and a noun.

element)

is

What follows

this VN ('ni') is a postposition indicating purpose or goal, and from that point of view

the verb form is functioning as

a noun on a par with the nouns in the following examples: (59) a. paatii ni iku 'go to a party' b. heya ni kuru 'come to a room' c. ronbun ni kakaru 'start on a paper' On

the

other

hand,

what

constructions is that the verb

is

peculiar

these

can appear with its argument

and complement (without the morphological modify a noun):

about

change on them to

103

(60} a. puuru de oyogi ni iku 'go swimming in the pool' b. kodomo to asobi ni kuru 'come to play with children' c. biru o kowasi ni kakaru 'start destroying a building' If the

VN form is really

a noun,

all the

complements and

arguments must be accompanied by the genitive marker, as has been mentioned previously, but that is not possible here: (61) a. *puuru de-no oyogi ni iku b. *kodomo to-no asobi ni kuru c. *biru no kowasi ni kakaru We

saw that

almost a

VN

compounds and

the

infinitive forms

complementary distribution

formation and nominal suffixation, that difference does not figure. (62) a. yama ni nobori ni iku mount.GO climb go b. yama-nobori ni iku

in regard

have

to compound

but in this construction

23

'go mountain-climbing'

(63) a. ueki ni mizu o yari ni iku 'go to water plants' plantDATwaterACCgive go b. ueki no mizu-yari ni iku (64) a. mado o huki ni kakaru 'start cleaning windows' wind.ACCclean start b. mado-huki ni kakaru 'start window-cleaning' SJ compounds are completely ambiguous, both verbal

and nominal usages

since they all have

(unlike native

VN's),

and

they allow both patterns: (65) a. gakkoo o/no sisatu ni iku

'go inspect schools'

23 0n the other hand, the two (compound and phrase) cannot mix in one configuration, as shown: (i) a. mizuumi de oyogi ni iku. 'go swimming in the lake' b.*mizuumi de sakana-turi ni iku.'go fishing in the lake' lake LOC fish -catch go (cf. mizuumi de sakana o turi ni iku.) lake LOC fish ACC catch go

104 b. kekka o/no hookoku ni kuru 'come to report the result' c. heya o/no soozi ni kakaru 'start cleaning the room' In these particular constructions, it seems as though the VN's have two with its

To its

different faces.

arguments and modifiers as

right it is followed by a case no

doubt

that

the

left,

a verb,

it combines while

marker as a noun.

morphological

nominal form and the infinitive

ambiguity

(verbal)

to its There is

between

the

form enables this

ambiguity, but we should also note that we find this kind of phenomena also with some cases of suffixation.

It is likely

that the agglutinative mode of

concatenation in Japanese is

one factor behind such cases.

We

of suffixation

that involves

a phrase

yome ni nari-te 'bride-to-be'. this case combines

would be

to say

syntactically with

have seen above one case in an

example like

Another way of conceiving of

that the

complement 'yome

'nar-' (V),

in

turn

morphologically combines with a nominalization suffix.

Each

combination is well-formed in the language; as

a

whole this

construction

configuration of Japanese.

does

which,

ni'

it is just that

not fit

the

regular

We will discuss more examples in

the following chapters, but it is to be noted here that such variety in the distribution of the VN form of a verb casts a question as to

what is a genuine case

part of derivational morphology.

of nominalization as

105 ~

3.

note on the condition on rendaku

In the preceding

section we have seen

numerous examples

of compounds in which the initial voiceless consonant of the second element compound

is changed to

formation.

a voiced consonant

This

phenomenon,

under the

referred

to

as

'rendaku', has been long noted, although the exact condition of its application has not been clarified. been claimed by those who details that

it is

(Martin 1952, section

investigated the process in great

largely unsystematic

Vance 1979).

that

Actually, it has

the

and idiosyncratic

It will be pointed out in this

consideration

of

morphological concatenations can give

the

different

us some insights into

the nature of this process. There

are two

generalizations over

the application

of

this process that have been attested to hold.

One is known

as 'Lyman's Law',

although the

observation

is

attributed to Lyman (1894), claimed

Motoori in the 18th C, Law

says

contained blocked.

that

anywhere For

(nama-gaki'raw (ai-kagi). a few

if

to go

further

to

Norinaga

according to Vance (1979).

Lyman's

there already in

the

second

instance, oyster',

a

voiced

element,

kaki'oyster' while

consonant rendaku

gets

kagi'key'

is

rendaku

never

does

This condition is almost exceptionless, and only

counterexamples have been

well-known generalization in dvandva

is

back

compounds.

found so far.

is the non-occurrence It is

The other of rendaku

this generalization

that we

106 will take a closer look [+N] can form

Three lexical categories of

here.

dvandva compounds,

and they

are exemplified

below along with their ordinary compound counterparts. (66)

N-N: oya-ko 'parent and child' (cf. sato-go 'foster child' mai-go 'lost child') eda-ha 'branches and leaves' (cf. waka-ba 'young leaves' futa-ba 'seed (lit. two) leaves') V-V: yuki-kaeri 'coming and going' (cf. Amerika-gaeri 'returnee from America' hi-gaeri 'day trip (lit. return)') yomi-kaki 'read and write' (cf. te-gaki 'hand written' hasiri-gaki 'quick-written')

AN-AN: siro-kuro 'black and white' (cf. iro-guro 'dark (lit. color-black)') What

distinguishes

dvandva

compounds is the lack of two

elements.

compounds

from

the

other

modifying relationship between the

Ordinary

compounds

are

generally

right-headed, with the first element modifying or specifying the second element in some way.

The accent pattern roughly

corresponds to this difference, although the accent patterns for

different

Namely,

dvandva

types

of compounds

compounds retain the

element and erase the rest, accent

of

yomi-kaki vs.

the

vary

second te-gaki.

subtle

accent of

ways.

the first

while in ordinary compounds the

element We

rendaku reflects the modifying elements in the compound:

in

just

dominates

are led

the

pattern:

to hypothesize

that

relationship between the two as the accent pattern marks

107

the head of the ordinary compounds (X Y),

the voicing marks

the second element as the head. In

addition

to

this well-discussed

application

of

rendaku

just

significant

tendencies we

significant contrast can

discussed,

find

with

element

of

there

these

on

the

are

two

this process.

be found among the

deverbal nominals (that were discussed second

condition

One

compounds with

in Section 2).

compounds

resist

The

rendaku

considerably when the first element is

a noun and stands in

DO relation

second element,

to the

base verb

opposed to when the first a

noun with

oblique

of the

as

element is an adjective/adverb or

relations.

The following

examples

illustrate this observation. (67) a. e-kaki (picture-paint) 'painter' te-gaki (hand-paint) 'hand painted' sita-gaki (under-write) 'draft' b. mesi-taki (rice-cook) 'rice cooking' mizu-daki {water-cook) 'casserole' kara-daki (empty-cook) 'putting an empty pan on fire' c. mono-hosi (thing-dry) 'place for drying laundry' kage-bosi (shade-dry) 'drying in the shade' d. syakkin-tori (loan-collect) 'loan collector' yoko-dori (side-take) 'snatch' e. sakana-turi (fish-catch) 'fishing' iso-zuri (beach-fishing) 'fishing on the beach' The

contrast in

the relationship

these compounds can

of the

be stated as the

two elements

following.

in

While DO

and a verb stand in argument-predicate relationship and they together name an action, an oblique noun or adjective/adverb

108 and a

verb stand

in modification

certain kind of act is restricted seems possible contrast

between

compounds

to its subkind.

case as

two elements

relationship,

Thus it

about the

ordinary

Namely,

well.

where the

in modification

and

which a

24

hypothesis we made

dvandva compounds

to this

readily applies stand

to extend the

relationship by

nominal

rendaku

in a

where

most

complex word 'modification'

excludes the predicate-argument relationship. There

is

another

verbal/adjectival compounds. all belong

N,

AN

and the rest;

There are

VN,

exists between

the former

greater extent than the

with

compounds formed

in which case the first element

(adjective stem),

interesting contrast

found

The compounds discussed so far

to [+N] category.

with the [+V] head as well, can be

contrast

significant

or

a prefix.

VN-V I

An

AN-A compounds

resist rendaku to a significantly latter.

Especially,

the Japanese

lexicon abounds in X-V compounds (unlike English), and it is

24 What remains a puzzle here is that the same contrast cannot clearly be found with SU-VN compounds, which should behave parallel to DO- VN compounds. Although there are some cases with this contrast (i.e., ame-huri 'rainfall' I hon-buri 'heavy rainfall'), many SU-VN compounds seem to show rendaku (cf. yuki-doke (snow-melt), hi-gure (sun-set)). Further, when the second element of a DO-VN compound consists of three syllables, it shows rendaku more frequently. So, although we have a contrast such as mahoo-tukai (magic-user, witch), hebi-tukai (snake-dancer) I hasiri-zukai (run-use 'errand person'), we also have vama-biraki (mountain-open), hito-qorosi (kill-person)'murder', and so on. It is more than likely that the fact that the first syllable of the second element in these compounds is accented (yama-bllraki) as opposed to the ones in (67) above (monotEo~) is one factor behind this difference. These problems rema1n for future research.

109

with these compounds that the following gives

contrast is most clear.

examples of the

fairly productive verbs and one

compounds formed adjective.

for only a few representative ones, rather

subtle

meanings

that

The

with two

(Gloss is given

since many of them have

would

take

many

words

to

translate,) (68) a. tatu 'stand' 'leave, depart' VN-V: ikiri-tatu, uki-tatu, omoi-tatu, ori-tatu, kiri-tatu, sosori-tatu, takeri-tatu, tut-tatu, ture-datu, tobi-tatu, nari-tatu, ni-tatu, nie-tatu, hiki-tatu, hurui-tatu, yuki-tatu, moe-tatu 'flare up', waki-tatu 'boil'. N-V:

awa-datu 'foam', ukiasi-datu, ozike-datu, ozoke-datu, ornate-datu, kasira-datu, kado-datu, kiwa-datu, keba-datu, saki-datu, sakki-datu, su-datu, sooke-datu, tabi-datu 'leave on a trip' tuna-datu, tunome-datu, tubu-datu, tuma-datu, tumasaki-datu, toge-datu, nami-datu, hara-datu, hi-datu, hukure-datu, mimi-datu.

AN-V: ara-datu, omo-datu, saka-datu. b. kiru 'cut' 'stop' 'do x entirely' VN-V: ii-kiru, uti-kiru 'put an end' osi-kiru omoi-kiru, kai-kiru, kasi-kiru, kak-kiru kari-kiru, kui-kiru 'bite off', si-kiru sime-kiru, sumi-kiru, suri-kiru, dasi-kiru tati-kiru, tate-kiru, tuki-kiru, tume-kiru de-kiru, nigari-kiru, nezi-kiru, nori-kiru hasami-kiru, hari-kiru, humi-kiru, huri-kiru mi-kiru, moti-kiru, yaki-kiru, wari-kiru N-V:

ura-giru, ku-giru 'divide up' se-giru, ne-giru ma-giru, yoko-giru 'cut across'

c. kurusii (A) 'hard' 'difficult' 'suffocating' AN-A: atu-kurusii 'hot and stifling' omo-kurusii kata-kurusii, sema-kurusii, musa-kurusii N-A:

iki-gurusii, muna-gurusii, kokoro-gurusii,

VN-A: kiki-gurusii, mi-gurusii, ne-gurusii

110 Nishio and

Miyajima (1971)

verbal and

adjectival compounds.

cases where

rendaku takes place

the following figures; number of

capable of

a list of

in each category

of rendaku,

to a

number of gives us

after the slash are the

conditions such changed

frequently used

Counting the

the figures

the potential cases

other phonological being

has

as the

voiced

namely,

where

first consonant

consonant and

the

Lyman's Law discussed above are met. VN-V compounds: 11/667 N-V compounds: 142/154 AN-V compounds: 10/10 The figures for First,

it

prefix-A;

adjectival compounds are not

seems hard

to distinguish

for instance,

as clear cut.

AN-A compounds

from

usu- 'thin' in usu-gurai 'slightly

dark' seems

to be

adjective,

especially since

preserved.

Secondly, there is one adjective kusai 'stinky'

which resists with

nouns to

translated as

functioning as a the

prefix rather lexical

rendaku in general, form

a compound

but is with

'smelling like x'.

AN-A compounds: 6/13 VN-A compounds: 9/9 (-kusai: 0/1) N-A compounds: 43/47 (-kusai: 2/25) Prefix-A: 11/12

meaning is

not

very productive

the meaning

We will list

separately below.

than an

roughly this item

111

The

correlation we

here:

find with

adjectives

if the first element of

frequency of

rendaku is very

is only

one-way

X-A compound is not AN, high (except for

the

the lexical

exception of -kusai). So far

we have put

tends to apply

forward the hypothesis

to compounds in which the

modification relationship. we

can

extend

mentioned.

it

to

that rendaku

elements stand in

It is not immediately clear how account

for

this

contrast

just

If we examine the headedness of these compounds,

however,

we notice

compounds we have

some

we see

compounded on its

points.

seen are clearly right-headed

the dvandva compounds. compounds,

interesting

The

[+N]

except for

Now looking at the verbal/adjectival

that when

a

verb or

left with a morpheme of

than its own (including prefix), clearly right-headed in

an adjective

a category other

the resulting compound

that it is the

is

second element that

determines the lexical category of the word as a whole. the other hand,

is

VN-V compounds and AN-A

On

compounds are not

clearly right-headed in this sense, since neither of the two items can

be regarded

whole word.

as determining

the category

of the

In terms of semantics, on the other hand, it is

not clear that they are non-headed (or hi-headed) as dvandva compounds because and X'.

In fact,

semantics vary: tobi-haneru

they cannot always among it can

'jump and

be paraphrased

by 'X

the numerous VN-V compounds, be a

hop');

conjunction 'V the

second

and V'

element can

the (ex. be

112 subordinate (ex. can

be

moe-tatu 'flare up'); or the first element

subordinate and

'exchange'); determine

the

in many

semantic

then,

speculate,

tori-kaeru

cases it head.

the

fact

seems

plausible

of (clear)

that

these

difficult to to

headedness is

compounds

resist

The contrast between DO and oblique nouns that we

rendaku.

found with deverbal

nominal compounds above is

spot with N-V compounds,

but

of N-V compounds not showing

rendaku (while 143 N-V compounds do),

v.

not easy to

it is rather significant that

among the 12 'exception' cases

DO and

(take-change)

is quite It

that this lack

for

responsible

(ex.

On the other hand,

9 of them consist of

many N-V

compounds with DO

show rendaku. (Also see the following footnote.) The observations made in this section can be summed up as the following.

Rendaku applies most

XY if X modifies Y and Y head of the compound. in

contrast with

is the semantic as well as lexical

The notion 'modification' here stands

'argument-predicate' relation

that of DO and V.

Rendaku,

marking

of

the

conceivable

readily to a compound

head that by

a

then,

complex

marking the

especially

can be seen as a way of word. head

It

is

further

and reflecting

the

modification relationship between the morphemes that make up a complex word, rendaku actually has the function of marking the

complex word

as a

lexical

language where phrasal and

and semantic

unit,

in

a

morphological concatenations are

113

sometimes obscured

in form.

such as

rendaku is

Japanese,

25

In an

agglutinative language

one of the

processes triggered by word formation. surprising well

that we

as motivations

find some in the

few phonological

It is therefore not

morphological conditions application

of this

as

process,

despite its often-remarked idiosyncracy to some extent.

25 In this respect the tendency found with N-V compounds that do not show rendaku is interesting. Namely, many of them seem to be in between a word and a N-V sequence with a case marking dropped, and do not seem to form a very tight unit as a word compared to the other N-V compounds. They are typically compositional and transparent in their meanings. For instance, compare kosi-kakeru (buttock-put) 'sit' (same in meaning as kosi Q kakeru) with te-gakeru (hand-put) 'start or try--ra project)•, or tosi-toru (age-take) 'to age' (same in meaning as tosi Q toru with ke-doru (sense-take) 'notice'. See Kageyama (1977) for the discussion on 'post-cyclic' noun incorporation versus 'pre-cyclic' noun incorporation.

CHAPTER III ARGUMENT STRUCTURE AND DERIVATIONAL MORPHOLOGY OF ADJECTIVES will

chapter

This

examine

involving adjectives as base.

argument structure of verbs

regulating the formation

deverbal

Although

nominals.

morphology

We discussed in the previous

chapter the important role that plays in

derivational

and the

distribution of

argument

structure

adjectives has

attracted less attention compared

of

will

verbs,

capturing

we the

see below

regularities

morphological derivation, there

a

is

nominalization verbs),

that

it turns

it

adjectives

Japanese

and

in

since

transparent

adjectives

out that the examination

for

exhibit

In fact,

productive

for

with that

is essential

as we will see.

completely suffix

that

of

(unlike

of deadjectival

nominals can yield some fruitful results for determining the argument structure of different types surface constituent

structure of

of adjectives and the

the sentences

containing

them. In

section

structure followed

of by

reflected in

1

general

problems

adjectives

in

English

discussion

on

how

will

argument

the nominalization and the

114

regarding be

argument discussed,

structure

is

compound formation

115 of adjectives. nominalizing point out

Section 2 we will discuss

In

suffixes for

Japanese

adjectives,

some theoretical consequences of

made about

them.

Section 3

two kinds of

will examine

and

will

the observation the verbalizing

suffix -garu, which has some interesting properties. 1. Argument structure of English adjectives 1.0.

English adjectives have not

detail

in terms It

verbs.

is

of their

argument

because

many

been discussed in much structure compared

adjectives

are

to

basically

one-place predicates (Theme, A)

and when they subcategorize

for a complement,

form of PP with a lexically

it takes the

specified P (ex.

[NP, of NP], [NP, to NP]), thus A', unlike

V', does not contain NP. On

the other

adjectives,

as

hand,

1

the

opposed to

basic

intransitive nature

verbs,

subcategorize only for PP's in English) by Riemsdijk

(1980)

from

grammar.

Namely,

generally

subcategorize

framework

in

the

adjectives ([+N]) points

out that

the point

the fact that for

assumption do not German

that they

(i.e.,

of can

has been questioned of view

of universal

English adjectives do not

NP's of

is the

assign case. adjectives

reflected case

in

theory

GB that

Riemsdijk (1980)

can subcategorize

for

1 There are three exceptions, as noted in McCawley (1982): worth, like, and near. See Maling (1980) for discussion on the synchronic status of them. It is claimed by Maling that while near passes most of the tests for being an adjective, the other two words behave more like prepositions than adjectives.

116 dative or genitive marked NP's, must be this

allowed to have

effect

he

and thus German adjectives

the case-assigning

proposes

a

property.

neutralization

of

To

syntactic

features ([+V, +N] -> [+V]) for German. It is thus how

the

plausible to state here that

generalizations

over case

structure of adjectives should be

the question of

pattern

and

argument

captured is still an open

one in universal grammar. 1.1. Complement inheritance under nominalization A

number

of

English

complements under

adjectives

nominalization.

can

preserve

The following

their

are some

examples: (1) eager to succeed ---> eagerness to succeed distinct from x ---> distinctness from x familiar with x ---> familiarity with x kind to Mary ---> kindness to Mary important to x ---> importance to x It has been

pointed out (cf.

cases (among others) functional operation

phrase-building.

(1983))

that these

present a problem to the assumption of

composition where

Moortgat

in

the

framework

word-building

of

always

syntactic precedes

Namely, although the nominalizing suffixes

such as -ness and -ity attach

to an adjective alone,

their

semantic scope in these cases clearly includes the entire A' (eager

to

succeed,

etc.).

Rather

than

loosening

the

117 constraints

on

the

syntactic operation

war~

of

grammar

(i.e., the strict lexicalist restriction), Moortgat opts for a

reworking

expressions

of

the

so that

ensured in the

semantics

the proper

semantics.

that

accompany

scope of

We are not

the suffixes

for this

type of

whether

there

consequences

are:

what kind of

Rather, what we restrictions exist

'complement inheritance' is

any

between

inheritance' and

difference the

phenomena;

in

postulation

that of phrasal

is

concerned here with

the technical details of how this is done. will be questioning

these

the of

and

empirical 'complement

suffixation that

we have

already proposed. We

will

first

consider the

generalizations

over

the

correspondence between the following expressions:

(2) a. X is A PP / to VP b. X's A-suff.(N) PP/ to VP Although the correspondence between (2a) be a fairly general one,

as

and (2b)

exemplified in (1),

completely so, as illustrated below: (3) a. John is familiar with the problem. John's familiarity with the problem. b. The problem is familiar to John. *the problem's famTliarity to John. (4) a. The job was easy for John. *the job's easiness for John. (5) a. The car was precious to Mary. *the car's preciousness to Mary.

seems to it is not

118

We can see that it is not the subcategorization features nor the particular suffixes

that rule out the

particular cases

of nominalizations such as the ones above. It has been noted since that some verbs (especially notundergo

early days of generative grammar 'psychological' predicates)

do

regular nominalization but rather what might be

described as 'passive' nominalization: (6) a. John amused Mary. b. *John's amusement of Mary. c. Mary's amusement at John. (cf. Mary was amused at John.) Based on (1980)

this fact

and other

argues against

cross-categorial

similar cases,

the assumption

correspondence

in

X' theory

between S and

expressed using grammatical functions only. these verbs

expressing

NP

that

can

be

She notes that

that behave 'irregularly' ·under nominalization

have NP's with position,

Amritavalli

the thematic functions of

and proposes

Experiencer as DO

that argument structure be

cross-categorial

correspondence.

Amritavalli

illustrates this claim by a triangle diagram of V, N, with their respective argument

v

(Causer, Experiencer) SU NP DO NP N

(Experiencer, X) Poss.NP PP

and A

structures and corresponding

subcategorizations. (7)

used in

A

(Experiencer, X) SU NP PP

119 This

diagram

is

directions of

supposed

represent

three

derivations as exemplified by:

amused(A), amusement(N); terror(N)

to

sad(A)

-> terrify(V),

derivational relations point Amritavalli

between these

(1980)

amuse(V)

-> sadden(V),

terrified(A).

tries

different ->

sadness(N);

There exist other

categories,

to make

but

is that

the

argument

structure of the verb is not paralllel to that of N or A for these triples,

s

between

and thus the cross-categorial correspondence

and NP

based on

subcategorzations breaks

down.

What is interesting for our discussion here is the fact that the correspondence diagram reflects

between A a type of

and N

as represented

nominalization under

complements of adjectives are inherited. it seems take

to be

its theme

argument

as

PP,

nominalization.

generally the case

which the

On the other hand,

that when

an adjective

argument

·as

subject and

the

this

PP

cannot

preserved

Take the

in the

be

contrast shown

experiencer under

in (4)

above,

repeated here: (8) a. John is familiar with the problem. John's familiarity with the problem. b. The problem is familiar to John.

*

The problem's familiarity to John.

In terms of subcategorization, both (Ba) and (Bb) PP], but in their argument structure (Ba) has [Exp. while (Bb) has [Theme, Exp.]. the PP

is obligatory,

Further,

while for

(Bb)

take [NP, Theme],

note that for (Sa) can be absent

in a

120 sentence

in

which

case

Experiencer is implied.

an

understood

or

arbitrary

This is partially similar

to the

realization of Agent/Theme arguments with derived nominals. (9) a. the enemy's destruction *(of the city) b. the city's destruction (by the enemy) As pointed

out in

Roeper {1983),

invoke the implicit agent (cf. 9a).

Similarly,

we

adjective

can

imply

for (4a,b) an implicit

experiencer cannot invoke an between (4) present under can.

(7)

and

out that in that sense the

is more central than

can say

argument can

9b), but not vice versa (cf.

It can further be pointed

genitive NP position

the theme

the PP position.

that the

theme of

an

while

an

experiencer,

implicit theme.

is that the Experiencer

nominalization (cf.(4b))

What differs PP cannot be

while the

Agent PP

Why there exists such a difference is not clear to me

at this point, but the facts indicate that for nominalizaton of adjectives

there is a

hierarchy of

Experiencer > Theme > others. that the

genitive NP

nominalization

is more

construction,

thematic relations:

Coupled with the assumption central than we

can

the PP

predict

thematic functions can appear in the following way. (10)

NP's

A-suff.

Exp. Theme *[ Theme

*[ Oblique

PP Theme

oblique Exp Theme

in the

that

the

121 This predicts the unacceptability of (3b), (4b) and (Sb)

as

opposed to (3a) as well as the phrases in (1) above. 2 Besides this problem inherited,

of what kind of

thematic functions

choice of the adjectives.

seem to

genitive NP in the

complements can be also regulate

the

nominalization of certain

Consider the following:

(11) a. John was sad. John's sadness b. The movie was sad. *the movie's sadness (12) a. John is cold. *John's coldness (not the personality sense) b. the room is cold. the room's coldness What is

puzzling here is that

in (11)

cannot be realized by the genitive experiencer,

while in (12)

the

the

theme argument

phrase as opposed to the situation is the opposite.

One possibility is that although these adjectives occur with NP's of more

than one thematic functions

as their subject,

one thematic function is more basic than the other, With sad

nominalization disambiguates this. see that the experiencer is more one thing,

the use of sad in

and the

it is easy to

basic than the theme;

(llb)

for

does not extend to its

2 Here we are ignoring the problem of how to assign Although it is a thematic functions to each NP and PP. serious and important problem that has been noted by many, it is beyond our concern in this study.

122 ~,

antonyms

glad,

the opposite pattern, adjectives it is

and with this and

not clear which,

functions is more basic. sense used in cold in

pattern of It is

(llb)

(12b)

On the other hand,

etc.

other temperature

either,

does not,

but how

nominalization as

cannot predict this pattern,

experiencer while

that is related

shown here

that mere

use of argument

of thematic

One can also note that sad in the

implies an implicit

at least clear

to make

if

cold shows

to the

remains unsolved.

subcategorization features

we would very likely need

and

structure in accounting

for these

facts. Let

us

go

inheritance.

back We

to

the

have seen

discussion

that what

preserved under nominalization is thematic

functions

they

of

complement

complements can

be

determined largely by the

carry.

It

is

not,

however,

determined solely by the specific suffixes, contrary to what is stated in

Moortgat (1981),

where he

classifies affixes

into two classes according to their transparency in terms of complement inheritance. seen here

Another fact about the data we have

is that the suffixes

belong to different levels in morphology.

that figure in

these cases

the level-ordered approach to

Namely, -ness belong to level 2, while -ity and

-ance belong to level 1. transparency and

This implies that the phonological

idiosyncracy are not relevant

of the transparent suffixes.

This

as property

is one feature that may

distinguish the cases of phrasal suffixation from complement

123 inheritance cases.

Phrasal suffixation, as conceived here,

involves only a specific set of suffixes that are productive and

show li-ttle

idiosyncracy.

The

cases of

complement

inheritance found with adjective nominalization do not this property.

sha~e

We will discuss further differences. between

the two later in this chapter. 1.2. Adjectival compounds So

far we

have

structure of

the importance

adjectives in accounting

nominalization. we also

discussed

Turning our eyes

find that argument

for the

claimed

in

Lieber

categories of N, A,

v,

and P,

that

NA,

AN,

AA)

the

the latter two (V, A)

not restricted in

P)

and A

any way,

it

lexical

are not.

the compounds formed with N

are

a crucial

Specifically, among

argument-taking while the former two (N, thus predicts that

facts about

structure is playing

(1983)

argument

to adjectival compounds,

role, contrary to what has been claimed. is

of

are She (NN,

since the

argument linking principle (ALP) does not apply (see Chapter 2 for the description of ALP.) compounds

with deverbal

Lieber (1983) be ruled

This is in contrast with the

elements,

to

which ALP

applies.

says that a compound of this type 'will never

out on

the ground

that one

satisfy its argument structure.' This however,

only

for the use

compound

structure.

of its

remark seems to hold,

of the one-place

Consider

the

stems cannot

adjectives in

following examples

possible NA compounds given in Lieber (1983):

of

124 (13} sky-blue, girl-crazy, color-blind, branch-brown, wife-blind, gun-crazy We

immediately

compounds. a

notice

that

there are

'A

like

N'.

The

examples

branch-brown above are of this type, razor-sharp, paper-thin, etc. color-blind,

gun-crazy,

compounds

as

NA

here is not that

argument-predicate.

and girl-crazy

'object' (in a broad sense) this PP that they form a

as well as

light-sensitive,

The relationship of N

of modification but rather

thin in that they

PP,

and church-goer are formed.

can take

and it is with the noun of

compound.

the way verbal

that of

adjectives such as sensitive,

differ from blue and

parallel to

above,

consider

user-friendly (as seen in a

and so on.

Namely,

and

and so are pitch-dark,

class-conscious,

computer advertisement),

and crazy

of

sky-blue

On the other hand,

water-resistant, blood-thirsty,

and A

types

One type has a noun modifying the adjective with

meaning

such

two

This seems to be quite

compounds such Further,

as truck-driver

we understand these

adjectival compounds not in any arbitrary way, but we relate them to the appropriate adjectival

phrase such as conscious

of class,

so on.

crazy about girls,

compound can

be formed

and

based on

Similarly a new

adjectival phrases--for

instance, see the following line taken from a newspaper. (14) Rain pounded the water-weary Midwest on this day (Chicago Tribune, June, 1984.}

125 Here again, water is the oblique argument of weary (weary of water)

and this information about the argument structure of

weary is crucial in understanding this new compound. in a way,

adjectives also

sister element in A'.

form

Thus,

compounds with their first

(See Chapter II for the discussion of

the first sister principle on verbal compounds.)

Note that

for the compound girl-crazy, it is not possible to assign an interpretaton that 'girl are crazy about x'. The other type, namely the modification type NA compounds ('A like

N')

since in an

can be

said to be

obeying the FSP

adjectival phrase blue like the

sky can be regarded

as the first sister to A,

the verbal compound hand-painted, is the first sister element. the modification

sky,

type NA

Q¥

as well, like the

just as for

hand of paint Q¥ hand

This observation predicts that

compounds contain

only one-place

adjectives, and this prediction seems to be borne out. compound girl-crazy

can either

mean crazy

about girls

The or

crazy like girls, but not crazy about something like girls. We can

conclude by saying

(1983) claims,

that contrary to

what Lieber

the possible formation and interpretation of

NA compounds are restricted by the argument structure of the adjectives, tendency

although this fact is often obscured due to the

for the

predicates.

majority of

adjectives

to be

one-place

126 2. Nominalization of Japanese adjectives

2.0. There are two morphological classes of adjectives in Japanese, often referred to as adjectives (A) and adjectival nouns (AN).

ending with -i in

Adjectives,

the predicate

position, inflect for tense and mood, while adjectival nouns take a copula:

-na for the prenominal position, -da for the

predicate position, and -ni for the adverbial form. not

be concerned

here about

since

classes, particular

they behave

derivations

chapter.

The term

we

the difference in

the

will

be

We will

of these

same way

under

discussing

in

'adjective' will henceforth be

two the this

used to

cover both classes. We

will

suffixes ,

first one

our typology.

discuss

two

types

of

of which will be argued to Then,

associated with the Japanese adjectives,

after

nominalization be syntactic in

reviewing the general problems

argument structure and case

marking of

we will use this syntactic suffixation

as a test to clarify some issues. 2.1. Two types of nominalizing suffixes In this section we will take up two nominalizing suffixes for Japanese adjectives and discuss their differences.

The

following are some examples of the adjectives that take both suffixes: (15) takai 'high' taka-mi I taka-sa hukai 'deep' huka-mi I huka-sa akarui 'light' akaru-mi lakaru-sa omoi 'heavy' omo-mi I omo-sa tuyoi 'strong' tuyo-mi ltuyo-sa sinken 'serious' sinken-mi I sinken-sa

127 The suffix -mi attaches to about 30 adjectives,

if we count

the words that are commonly used, while the other suffix -sa seems

to be

completely

difference between abstract

state

or

concrete

element

productive.

the two forms property

is that A-sa

while

bearing that

The basic

A-mi

semantic

denotes the

denotes

property.

The

a

more

following

examples illustrate this point:

(16) a. Kawa no huka-mi I *huka-sa ni hamat-ta. river GEN depth LOC fall-PAST '(I) fell into the deep point of the river.' b. Kawa no *huka-mi I huka-sa ni odoroku. river GEN depth LOC be surprised '(I) am surprised by how deep the river is.' (17) a. Yuka ga hon no omo-mi I *omo-sa de hekomu. floorNOMbook GEN heavy-SUF CAUSE dent 'The floor got dented due to the weight _of the books.' b. Hon no *omo-mi I omo-sa o hakaru. bookGEN weight ACC measure '(I) measure the weight of the books.' Thus,

while A-sa has a transparent meaning,

A-mi is

Roughly speaking,

for huka-mi

taka-mi 'high place' and so on,

it denotes a

not predictable.

'deep point',

the meaning of

particular point or place with that property; for yowa-mi 'a weakness', omo-mi '(heavy) weight', and so on, it denotes an entity with that property. These

observations

lead

us

to

believe

that

suffixation should be part of the lexicon, while -sa is not. In the level-ordered morphology, that is, the former belongs to class

l suffix while

the latter

to class 2

or higher.

128 This speculation

is further confirmed

adjectival compounds can feed the of them

can feed the

by the fact

that no

-mi suffixation while all

-sa suffixation.

See

the following

examples. (17) hono-akarui --> hono-akaru-sa /*-mi (cf. akaru-mi) 'slightly light' muzu-gayui --> muzu-gayu-sa /*-mi (cf. kayu-mi) 'tickling and itchy' ama-zuppai --> ama-zuppa-sa /*-mi (cf. suppa-mi) 'sweet and sour' soko-hukai --> soko-huka-sa /*-mi (cf. huka-mi) 'bottom-deep, deep-bottomed' Another property of the -sa suffixation that is not found with the -mi

suffixation is that there exists

a pattern of

correspondence between S and the nominalization: (18) X ga A. 'X is A.' -->X no A-sa 'X's A-ness' This correspondence

is especially clear

when it

is acting

semantically as a clause nominalizer: (19) Taroo wa Hanako no kasiko-sa ni kizuk-anai. TOP GEN clever-ness LOC notice-NEG 'Taro does not notice how clever Hanako is.' On the other hand,

some cases of the -mi suffixation cannot

be accompanied by the genitive NP: (20) a. hukami no aru midori 'deep green' depth GEN have green b. *midori no hukami 'green's depth' We

therefore

suffixes

belong

conclude to

that

different

these places

two in

nominalizing our

typology.

Namely, the -mi suffixation belongs to the lexical rule with morphological operation,

while the

-sa suffixation belongs

to the syntactic rule with morphological operation.

We will

129 see further evidence

below for the syntactic

nature of the

latter. Kageyama

(1982)

has

~

argued that

has

the

boundary(+), lower than the compound boundary (#i), phrasal boundary(#), to stems rather effect on

or the

given the facts that '-sa is attached

than full-fledged words and that

the accentuation of

this assumption

affix

he attempts

whole derived to explain

it has no

words.'

that a

With

productive

AN-forming SJ morpheme -teki (which has a compound boundary) cannot be nominalized with -sa, given the correctness of the level-ordering hypothesis. (21) *genzitu-teki-sa 'realistic-ness' *ippan-teki-sa 'general-ness' This claim, First

however,

of all,

creates more problems than it solves.

it cannot

explain

why -sa

compound adjectives, as shown in (17). that it does

the accent pattern either. above,

there

is an

As for

gensitu-sei,

and

the fact

can also follow

the facts shown in (21) that Kageyama

Namely, these examples are

existence of words with the ippan-sei,

to

which does not change

alternative explanation

himself suggests (and rejects). blocked by the

Secondly,

not affect the accent pattern

from assigning it a phrasal boundary,

can attach

so

on.'

SJ suffix -sei;

In fact

N-teki-sa

3 For discussion of 'blocking' of a certain morphologically derived form by the existence of another form, see Aronoff (1976). In this particular case, even though N-teki-sa should be well-formed due to the productivity of -sa, the form N-sei is preferred over it,

130 becomes more acceptable when there N-sei to that word honkaku-teki-sa to

the judgement

maintain that

reported by

is no corresponding form

L *honkaku-sei,

Kageyama

(1982).

the -sa suffixation belongs to

contrary We

thus

the syntactic

rules. 2.2. Argument structure and case marking of adjectives As

is the

case with

Japanese adjectives however, NP's

a fair

and

English adjectives,

a number

are one-place predicates.

number of adjectives which

they have

been

one

source of

controversy in Japanese syntax,

as

of

There are, occur with two

a

long-standing

we will see immediately

below. There

are so-called

'stative

predicates' in

Japanese,

which are characterized by taking the nominative (instead of accusative)

marking

case

on

their

'object'

NP.

Morphologically, many of them are adjectives, although a few verbs fall into marking patterns,

this group as well.

In terms

of the case

adjectives belonging to this class can be

classified as follows: (22) A. NP ni NP ga A. (ex. Boku ni kore ga omosiroi.) DAT NOM 'I am amused at this.' B. NP ni/ga NP ga A. (Boku ni/ga hebi ga kowai.) 'I am fearful of snakes.'

c.

NP ga NP ga A. (Taroo ga turi ga umai.) 'Taro is good at fishing.'

probably because it is simpler in form, and also, since the stem is a SJ morpheme, a SJ suffix is preferred over a native one.

131 The difference between classes A and B above, namely whether a certain adjective allows the marker

on

the

first

controversy

that

following.

One

nominative ordinary second

on

by

these

approach takes

the

1973)

to

as

the

takes the

basic (as

the

The nominative

marker on

the

the ga/ni

derives

is

basic

be

hand,

is derived

by the

o/ga

marker on the first

conversion.

In short

constructions

adjectives from the ordinary other

Kuno,

first NP

the other

The

adjectives

approach (e.g., the

subtle.

and the alternating dative

is derived

approach

rather these

subject marker).

conversion, NP

NP is

surrounds

marker on

NP,

nominative as well as dative

with

this

two-place

transitive construction. [NP ni

NP ga

A]

as the

The basic

pattern and derives the nominative marker on the first NP by In this approach this

a conversion rule. regarded

as basically

different [NP ga

construction is

from ordinary

Specifically,

constructions

that take

Kuroda (1978)

has called this pattern 'ergative'.

reasons to be discussed in this latter approach

and call

NP

o V].

section,

transitive

For the

we will adopt the

the constructions

of (22

A,

B)

'ergative'. ((22 C) will be discussed later.) With these remarks on with one

class of

the particular problems associated

adjectives,

we can

now list

the basic

classes of adjectives with their argument structure and case markings. (23) A. Theme ga A. (ex. Yama ga takai.) 'Mountains are high.'

132 B. Exp. ga A. (ex. Boku ga uresii.) I am glad.' I

c.

Exp. ni Theme ga A. (Boku ni hebi ga kowai.) I am fearful of snakes. I

I

D. Theme ga Goal ni A. (Taroo ga Hanako ni yasasii.) 'Taro is kind to Hanako.'

E. Theme ga Obj. ni A. (Taroo ga samusa ni yowai.) 'Taro is weak against the cold.

I

F. Theme ga [Theme ga A.] (Taroo ga se ga takai.)

'Taro is tall. ' (lit. T. is (height is tall).)

2.3. Complements under nominalization We will examine

in this section how

applies to the adjectives that and discuss what the result can

the -sa suffixation

take more than one argument, tell us about the structure

of sentences involving those adjectives. Let us first see what

happens with the nominalization of

adjectives as classified in (23) above. (24) A. Theme ga A. --> Theme no A-sa. (Yama no taka-sa)

B. Exp. ga A. --> Exp. no A-sa. (Boku no uresi-sa)

c.

Exp. ni Theme ga A. --> (*Exp. no) Theme no A-sa ((*Boku no) hebi no kowa-sa)

D. Theme ga Goal ni A. --> Theme no (*Goal no) A-sa (Taroo no (*Hanako no/ni) yasasi-sa)

E. Theme ga Obj. ni A. --> Theme no (*Obj. no) A-sa (Taroo no (*samusa no/ni) yowa-sa) F. Theme ga Theme ga A. --> Theme no Theme no A-sa (Taroo no se no taka-sa)

133

The thematic function hierarchy for the genitive NP position in nominalization is Oblique.

We also

the following: note that

Theme

Experiencer

>

the genitive

NP most

often

corresponds to the nominative marked NP (but see below). for the

other complements of

simply be English.

inherited, One

NP in

as opposed to

factor

restriction we have

the adjectives,

behind

Japanese cannot have a

saw with

difference

mentioned in the previous PP adjunct,

As

they cannot

the cases we

this

>

is

the

chapter that

but can

only be

modified by a genitive marked NP. Now

let

us

take

adjectives mentioned (22)

al:>ove,

the

a

closer

above.

look

at

Taking the

the

'ergative'

classification of

following illustrates

the

pattern

of

nominalization of these adjectives: (25) A. NP ni NP ga A. (*Baku no) eiga no omosiro-sa 'the movie's funniness (to me)' (*Baku no) kodomo no kawai-sa 'the child's dear-ness (to me)' B. NP ni/ga NP ga A. (*Boku no) gan no osorosi-sa 'cancer's scariness (to me)' (*Boku no) H. no urayamasi-sa 'H's envy-causing-ness (to me)' (*Baku no) sore no hazukasi-sa 'its embarassing-ness (to me)' (The class (22 C) will be discussed below.) the Experiencer argument can never NP,

even

though many

of them

We can see that

realized as the genitive

(i.e.,

those

belonging to

134 (22B)) can take the nominative

(instead of dative) marking.

This fact indicates that the nominalization pattern does not correspond to situation

the case

in

(25)

markings. above

is

Note in

further that

contrast

nominalization of the ordinary transitive verbs,

with

the the

where both

Agent and Patient can be realized by the genitive NP: (26) Ginkoo no doru no kaisime. bank GEN dollar GEN purchase 'the purchase of the dollar by the bank' The impossibility

of nominalizing

in this

fashion on

the

part of the ergative adjectives here is significant, because this indicates that,

at least in some respect,

plausible to treat these two-place ~ransitive-

the

verbs

case-markings (e.g.,

by

observe the nominalization of (22)

that cannot

adjectives on a par with

manipulating

by o/ga

genitive NP's

just

surface Now

the 'two-place' adjectives of

take the dative marking on

and the

the

and ga/ni conversions).

We notice that they fall into two groups; two

it is not

other

the first NP.

one that can take

that cannot

comfortably

nominalize at all: (27) a. Taroo no tenisu no uma-sa /heta-sa GEN tennis GEN good bad 'T's being good/bad at tennis' b. ??(Boku no) kuruma no hosi-sa I kai-ta-sa I GEN car GEN want buy-want 'my wanting(to buy) a car.' c. *Boku no Hanako no suki-sa /kirai-sa I GEN GEN like dislike 'my (non-)fondness of Hanako'

135 There

are

several

ways

in

which

these

different from the ones shown in (26) First of all,

Secondly,

is

Expe~iencer

and

a~ong

they cannot omit the first

elliptical. (27b,c)

adjectives

the first

NP without being

NP in

argument, the one in (27a)

This

conception

of

everything

seems

cannot

argument to

be

allowed

if

and is an

is a Theme argument.

we

take two Theme

under On

str~cture.

follow

(27a)

the latter

In other words adjectives of the class (22C) arguments.

themselves.

the difference between

that while

are

an

the

consider

ordinary

other

hand,

that

these

adjectives belong to the class (23F) above, namely that they make an A' "-'ith its i:nmediate theme NP,

which in turn takes

another theme N? to make S, as shown below: ( 28)

a.

b.

S

~

~

N?

~'?

,:

To,..-oo~o..

~

t-1?

'bad

at' are

usually

p...

-rw.." Exp. ga A-garu. (Taroo ga uresi-garu.

'Taro acts happy.'

b. Theme ga A. --> Exp. ga Theme o A-garu. (Taroo ga hon

omota-garu.

0

'Taro finds the book heavy. ' ) c. Exp.ni Th, ga A. --> Exp. ga Th o A-garu. (Taroo ga hebi We can

say that

0

kowa-garu.

-garu converts

a basically

pattern of adjectives (of feelings) pattern typical of verbs. -garu is sensitive

'T. fears snakes.')

into a transitive case

The claim that the suffixation of

to the argument structure

is further supported

ergative case

by the fact that if

of adjectives

a one-place Theme

adjective is not capable of optionally taking an Experiencer argument,

this

terms this means

suffixation does

not apply.

that adjectives that can

denote a property

that belongs to the

In semantic

occur with -garu

subjective feelings.

For instance, it seems that color terms can never occur with -garu.

(cf. *aka'red'-garu, *siro'white'-garu, etc.)

As noted

in Abe

(198lb),

derived adjectives.

-garu

For instance,

can attach

to various

there are

some fairly

productive A/V conversion patterns as exemplified below: (48) a. v-asi --> A

149

urayam I urayam-asii 'envious' wazurau I wazur-asii 'bothersome' b. A-mu --> V osii I osi-mu

'to miss'

tanosii I tanosi-mu 'enjoy' The suffix -garu

can attach to these adjectives

and add to

this pattern of conversion. (49)

a. NP ga NP o urayamu. 'NP envies NP. ' NP ni NP ga urayam-asii. NP ga NP o urayam-asi-garu.

b. NP ni NP ga osii. 'NP misses NP. ' NP ga NP o osi-mu. NP ga NP o osi-garu.

Further, it has been noted by many (e.g., Kuno 1973) that -garu

attaches

to

the desiderative

suffix

-tai

and

it

Kageyama 1982)

by

cancels the 'object-ga' marking induced by -tai: (50)

a. Taroo ga [mizu o nomi]-tai. 'T. wants to drink water.' b. Taroo ga mizu ga nomi-tai. c. Taroo ga mizu o nomi-ta-garu.

This fact has been accounted the

percolation

of

for (e.g.,

the feature

[+stative]

of

the

element, which triggers the 'object-ga' marking. (51) a.

[+stative]

no~:') [-st.]

[+st.]

b.[-~~ nom1

ta

garu·

[-st.] [+st.] [-st.]

head

150 In this

view the case

markings of (50a-c)

are determined

solely by the category of V under which the suffixations are On the

performed.

other hand,

we have

already seen some

evidence from the

way V'-tai are nominalized

the

suffix

desiderative

-tai basically

full details in Chapter IV.) bracketing of (50a)

attaches to despite

attaches

(This position will be argued

category of V'.

(as the

with -sa that

that constituent,

its appearance

(SOc)

to

the

for in its

If we maintain that assumption shows), then is

and also we would

not

that -garu predict that

a truly

transitive

construction but rather an intransitive construction: (50)c 1 Taroo ga [[mizu o nomi]-ta]-garu. This prediction is

borne out,

as demonstrated

by the fact

that sentences of the form (SQ.c) cannot passivize: 5 (52) a. Taroo wa Hanako o sasoi-ta-gat-te iru. 'Taro wants to ask Hanako out.' b. *Hanako wa Taroo ni sasoi-ta-gar-are-te iru. Compare

(52)

predicate

with

with the

the

following,

morphologically

which similar

involves

the

derivational

history (V->A->V). (53) a. Taroo wa Hanako o urayam-asi-gat-te iru. 'Taro envies Hanako.' b. Hanako wa Taroo ni urayam-asi-gar-are-te iru.

5 This does not, of couse, preclude the possibility of adversative passivization, which can apply to intransitive verbs. (i) Taroo wa kodomo ni iede-si-ta-gar-are te komatta. TOP child DAT run away want PASS troubled 'T. is troubled because his child wants to run away.'

151 The derivation of the predicate in (53),

however,

does not

involve a V' category, so (53a) is a transitive construction as the possibility of passivization shows. significant,

especially in view

This contrast is

of the lexicalist analysis

proposed for the suffixation in (52) and (53),

which cannot

distinguish them in any principled way. Note

that the

standard

transformational analysis

Kuno 1973) with sentence embedding, at

EQUI,

the end of the cycle cannot naturally

unacceptability of (52b).

and verb raising for

account

the

It is because when -garu attaches

to the verb by verb raising in not a

(cf.

the last cycle,

constituent any more due

[NP o V] is

to the verb raising

in the

previous cycle:•

(54) a.

T ga

T ga

b.

T ga

T ga H o sasoi-ta]-garu.

T ga H o sasoi ]-ta]-garu.

This concludes the discussion of this suffix is

-garu.

We argued that

not a transitivizing suffix

but rather what

it does is convert the pattern.

It also

ergative pattern into the transitive

requires the Experiencer argument

the subject NP of the derived to the argument structure of Further,

we

verb,

to be

and hence is sensitive

the adjectives it attaches to.

have demonstrated that

-garu attaches

to the

6 rn order to block such ungrammatical sentences as (52b) Kuno (1978) assumes a Global Constraint, originally attributed to Harada, which prohibits the Passive from subjectivizing an NP that used to be a constituent of a sentence embedded in the sentence to which the rule applies.

152 V'-tai constituent. for the status

We will argue in the following chapter

of -tai as a phrasal suffix

consequences of that claim.

and discuss the

CHAPTER IV PHRASAL SUFFIXES I: ALTERNATING CASE MARKING In

this

predicates

we

in Japanese

marking', better

chapter

and

will

take

which allow

argue that various

explained

if

we

up

some

so-called 'objectfacts about them

assume

stative

that

some

g£

can be

suffixes

are

'flexible' in their subcategorization in that they basically attach to

V' category

attaching only to on the

but sometimes

the verb of V' as a

surface constituents.

reanalysis

that

are best

we

are

The

analyzed as

result of reanalysis

particular form

concerned with

here

is

of the as

the

following: (1) a.

X

v

]vt + suff. --->

b. [X] [ V + suff.]

1. Phrasal suffixation and reanalysis In the framework assumed here perceived as a syntactic in this case, of

type

A

suffixation to a phrase is

rule with morphological operation;

a phrase of a category V' is formed by a rule (syntactic

rule

by

syntactic

operation--see

Chapter I for the rule typology),

then the output is fed to

a

rule

rule

of

type

B

(syntactic

operation).

As

is the property

process (la)

above is regular,

153

with

of syntactic

morphological rules,

the

not constrained in terms of

154 idiosyncracies (see below for examples). mentioned in Chapter I, build a

phrasal suffixation can potentially

structure that is

rules.

As it was briefly

not derived by

phrase structure

Suppose the suffix has the feature [+N] and the verb

subcategorizes for a direct object NP. (2) [ NP-o

V ]v' + suff.(N) ---> [ NP-o [ V + suff.]N]

Such a sequence, NP is

a nominal predicate follwing an accusative

not generated by

type

of

special

discussion

Japanese but only

suffixation.

of this

similar case from (1980).

PSRs of

(See

particular

Chapter

type

V

for

the

of suffixation.)

A

Greenlandic Eskimo is reported

In Greenlandic noun

by this

incorporation,

by Sadock

morphemes can

incorporate a possessed NP, which is a syntactic phrase: (3) [ NP-erg. N] +verb---> [ NP-erg.] [ N-verb] Since a possessive NP gets

an ergative case,

sequence of an ergative NP

followed by an intransitive. verb

is not

observed in Greenlandic other

noun incorporation. property according

of

being

than in this

Sadock (1980) able

to Wasow's

to

build

the resulting

points out a

(1977) criteria,

new

type of that this

structure

is,

that of

syntactic

certain

cases

rules rather than lexical rules. It

can

consequently

be

said

that

of

syntactic word formation rule create a 'marked' structure in the language. structure

It is

building can

also conceivable result in

which is often susceptible to

that this

a structural

a reanalysis.

type of

ambiguity,

For instance,

155 suppose a verb of a category X assigned a certain case x for its object argument, while a verb of a category Y assigned a different case feature

y,

and there

existed a V'-suffix

with the

+V, +Y ], which can attach to a verb of a category

X: {4) [ NP-x This

V[+X] ] + suff.[+V,+Y]

suffixation

language;

namely,

results

in a

marked

an NP of case x

structure

in

the

followed by a verb of a

category Y: {5) NP-x V-suff.[+Y) It is conceivable

that the NP in {5)

argument of the complex verb as

get

reanalyzed as an

a whole and assigned a case

y instead:

{6) NP-y

v-suff.[+Y]]

by analogy to a more unmarked construction in the language. 2. Facts about Japanese stative predicate constructions It is well known that the predicates in Japanese with the feature

[+stative] assign

a nominative

'object NP' {henceforth 'second NP)

1

~

case

Also,

to

their

many of them can

mark their subject NP with dative case as well as nominative case.

Semantically

these

are

predicates

feelings, desire, non-intentional perception, so on.

Morphologically

they can be verb,

expressing ability,

and

adjective,

or

adjectival nouns:

1 See Section 6 for discussion on refer to them as 'object'.

why we do not

want to

156 (7) Taroo ga/ni hebi ga kowai. (adjective) NOM/DATsnake NOM fearful 'Taro is fearful of snakes.' (8) Taroo ga/ni eigo ga dekiru. (verb) NOM/DAT Eng.NOM can do 'Taro can speak English.' (9) Taroo ga/*ni Hanako ga suki-da. (adj. noun) NOM/DAT NOM like COP. 'Taro likes Hanako.'

It is also the case that

verbs suffixed with a desiderative

suffix -tai and potential

=(~)eru

mark their second NP with

lli!.:

(10) Taroo ga/ni eigo ga hanas-eru. NOM/DAT Eng.NOMspeak-can 'Taro can speak English.' (11) Taroo ga/*ni eigo ga hanasi-tai.

NOM/DAT Eng.NOM speak-want 'Taro wants to speak English.' The standard analysis for sentences ·like (7)-(11) assume

certain predicates

lexically

assigned

to

to bear

them and

the feature trigger

conversion of Q --> lli!. on the 'object' NP. When the [+stative] predicate is (10 )

and ( 11 ) ,

a

[+stative]

case

marking

(See Kuno 1973.)

a 'higher' predicate as in

this conversion takes place

on the higher

cycle: (12) a. [ Taroo ga eigo o hanasi] b. [Taroo ga [Taroo ga eigo o hanasi] -tail

.J. fl

is to

-1-

ga

c. Taroo ga eigo ga hanasi-tai.

157 In some

cases this

conversion is

conversion of 9£ --> ni on

followed by

the subject NP.

an optional

As shown above

some predicates (cf. 9, 11) do not allow this dative case on the subject

NP,

which presumably has

to be marked

on the

predicate as its idiosyncratic feature. It has been noted that

under certain circumstances there

is an alternative set of case markings as shown below: (13) Taroo ga eigo o hanas-eru. (14) Taroo ga eigo o hanasi-tai. In the accounts given so far, this fact can be dealt with by somehow preventing feature

of the

conversion

is

suggests that following

the feature

[+stative] from

whole

predicate,

blocked.

For

in

order

that

instance,

feature percolation

structure

so

the

Q -->

Kageyama

must be

for

the

becoming a

(1982}

blocked in alternating

9£

the case

marking pattern to be allowed.

-tai [+stative] The gQ --> ni conversion on

the subject NP does not take

place if the object is marked with Q: (15) *Taroo ni eigo o hanas-eru (koto}. The

ungrammaticality of

accounted

for by

every sentence must 1978.)

a

sentence

postulating

the

like (15)

has

been

general principle

that

have a nominative NP.

( See Shibatani

158 3. Proposal Although the account given is sufficient to generate both (11)-(12) and (13)-(14), it does not predict when we get one form rather than

the other at all.

As we

will see below,

the two sets of case markings of (11)-(12) and (13)-(14) are not interchangeable one

pattern

sometimes

in many instances.

over

rather

the subtle

different speakers.

other

with

and

also

The

this

preference of

construction

seems

to

vary

is with

Nevertheless, there seem to be several

factors that are relevant in the

choice of one pattern over

the other. What we will do in the alternating

case marking

V'-suffix as

a V-suffix

will be

argued that

relevant here fall

following is to propose that this is

due to

the

under certain

many of

reanalysis of

circumstances.

the factors

that seem

together if we consider them

a It

to be

as part of

the environment for this reanalysis. Using the sentences (11)-(14), we will now illustrate our hypothesis.

First,

we

assume that the suffixes

-reru basically subcategorize for a V'. transformational approach supposed to have the

to these

-tai and

In the traditional

constructions they

following underlying structures,

the suffixes being the sister to an S node.

are with

159

s

a.

( 16)

s

b.

~

Taroo

~

V'

Taroo

~

S

s

-tai

-eru

Taroo

V'

~

eigo

~

~

~

Taroo

V'

hanasi-

V'

eigo hanas-

However, when the suffixation takes place as a result of the predicate raising

on the

higher cycle,

actually sister to V', not S. -reru are subject

so-called 'Equi' NP's in

both

relevant structure when

the suffixes

are

This is because both -tai and predicates,

S's to

be

which

require the

coreferential.

So

the suffixation takes place

the

is the

following:' ( 17)

s

a.

b.

~V'

Taroo ga

Taroo ga

~

/~

v•

-tai

V'

/"'"'hanasi

eigo o The

assumption

supported V'

by the

that -tai fact that

V'

eigo o and

-reru

they can

-(r)eru

han as attach to

attach to

V'

is

conjoined

's:

'The mapping of (16) to (17) can presumably be accounted for by 'rules of LF' rather than by syntactic transformations, but the discussion here holds equally well for that approach.

160 (18) Taroo wa [gakkoo e itte, tomodati ni ai]-tai/-eru. TOP school GOAL go friends DAT see 'Taro wants to/can go to school and see friends.' but not to conjoined S's with different subject NP's: (19)

This

[ Taroo wa utai, Hanako wa odori ]-tai/-eru. --->*Taroo wa utai, Hanako wa odori-tai/odor-eru. TOPsing TOP dance 'Taro wants to sing, and Hanako wants to dance.' 'Taro can sing, and Hanako can sing.'

assumption

natural

when

about

we

V' suffixation

consider

the

seems

semantic

to

scope

be

also

of

these

suffixes. Going back now to (17), after the suffixation takes place we get the following set of strings:

(20) a.

Since

eigo o hanasi ]-tai ] ~anas

b.

eigo o

-tai

is lexically

adjective

preceded by

an

]-eru ] adjective

an accusative

NP,

the result

is

which does

conform to any phrase structure rules of Japanese.

above, which,

This

building discussed in Section 1

according to our assumption on syntactic word

formation, occurs only with phrasal affixation. although the suffix -reru is lexically a verb, [+stative],

not

There is

no adjective that subcategorizes for an accusative NP. is an instance of structure

an

and there

is no simple verb

Similarly, it is marked

with that feature

which subcategorizes for an accusative NP in Japanese. below for the discussion discussed in Section 2,

of apparent counterexamples.)

(See As

all simple [+stative] predicates in

Japanese mark their second NP

with~·

161 It is now easy to see that the environment for reanalysis is supplied in (20a,b).

All that is needed is rebracketing:

(21) a. Taroo ga

eigo ga

hanasi-tai] ]

b. Taroo ga

eigo ga

hanas-erul

The

NP

is

interpreted

predicate as a whole,

as

and

an

argument

of

the

stative

hence is assigned the nominative

case marking. Our

hypothesis is,

sentences (13)-(14).

(11)-(12)

then, is

that

a result

We will now see

the case

of

marking

of

reanalysis based

on

what kinds of predictions this

hypothesis can actually lead us to. 4. Morphological transparency Our hypothesis immediately predicts that in order for the alternating predicate,

case

marking

patterns

to

exist

for

one

the predicate must be morphologically complex in

order for the base for reanalysis noted in Kuno(1973:

95)

that

to exist.

It is briefly

the alternating case marking

exists only with 'stative derivatives', by which Kuno refers to

complex

predicates

involving

hypothesis naturally predicts that. predicates, which

can

however,

there exist

occasionally

nevertheless are

take

not generally

-tai

and

-reru.

Our

Besides -tai and -reru a few stative predicates accusative

considered as

NP's

which

derivatives.

These predicates are suki-da, kirai-da, hosii, and wakaru.

162 We will take up the

first three predicates first,

which

are given in the following examples: (22} Taroo wa Hanako ga/o suki-da/kirai-da. TOP NOM ACC like dislike 'Taro likes/dislikes Hanako.' (23} Boku wa kenna kuruma ga/o hosii. I TOP such car NOM ACC want 'I want a car like this one.' though

Even

these

are

constructions

not

generally

considered as involving a higher predicate, these predicates are

nonetheless

verbs which

morphologically related

mark their second

NP with an

to

corresponding

accusative case.

They are shown below: (24) Taroo wa Hanako o suku/kirau. 'Taro likes/dislikes Hanako.' (25} Boku wa kono kuruma o hossuru. 'I want this car.' It is then possible to say that when the second NP is marked with a

it is

nominative case,

an instance

of reanalysis

parallel to the cases with -tai and -reru: (26} Taroo ga [ Hanako o suk ] -i-da. ---> Taroo ga [ Hanako ga [ suki-da ] ]. (27} Boku ga [ kono kuruma o hos ] -ii. ---> Boku ga [ kono kuruma ga [hosii] ]. One difference between these cases and the ones involving higher predicates is that with

(26)

marking

preferred

on the

marking in most with these form.

second

NP is

contexts.

and (27)

So it might

predicates the reanalyzed

Note that the morphological

to an

a nominative accusative

be considered that

form is

the unmarked

derivation we have here

163 is not productive -reru. with

at all

3

like the suffixation

of -tai and

Hence it is easier for these forms to be lexicalized the reanalyzed

case

pattern.

(See Section

6

for

further discussion and theoretical implications.) There is that

an additional small

morphological

piece of

transparency

affects

evidence showing the

reanalysis.

Observe the following examples: {28) a. Boku wa Hanako ga/o suki-da. 'I like Hanako.' b. Boku wa Hanako ga/??o dai-suki-da. 'I like Hanako very much.' The emphatic prefix dai- attaches only to [+N] category. the prefixation interpretation

of dai- strengthens the of

the

morphological relation less acceptable can be

predicate

suki

to the verb less

status of the

adjectival nominal and

makes

transparent.

accusative marking

attributed to this morphological

So

the The

in (28b)

difference between

3 Besides the ones discussed above, there is one expression o-ki-ni-iri (A) 'favorite' derived from the verb expression ki ni iru (lit. 'suit one's feeling'): Tanaka-sensei wa Hanako ga/o o-ki-ni-iri da 'Hanako is Prof. T's favorite'. This expression is lexicalized and must be used with this honorific prefix, o-. Thus, generally speaking, although the nominal formation from a verb stem by adding -i is extremely common, it rarely forms an adjectival nominal. For example, from a verb nozomu 'to hope for' a noun nozomi 'a hope' can be derived, while an adjective nozomi (what would be a parallel form to suki) does not exist. On the other hand, the derivation--of hosii (Adj) from hossuru (V) is idiosyncratic in form, while a parallel derivation with a suffix -asii is semi-productive; ex. nozom-asii 'desirable'. This suffix, however, does not allow the case alternation as we discuss here; sore ga/*o nozomasii 'that is desirable'. Thus this affix attaches only to V, but not to V'. For discussion of this suffix and others relating adjective and verb, see Chapter III.

164 the two predicates. We

should now

predicate wakaru, and

~-marked

consider the

exceptional

nature of

the

which can be preceded by both Q-marked NP

NP:

(29) a. Boku ga/ni kimi no kimoti ga wakaru. I NOM/DAT you POSS feelings NOM understand. b. Boku ga kimi no kimoti o wakaru (koto). 'I understand your feelings.' This predicate cannot be given a parallel analysis, since it is

a

lexical verb

and

not

a

derivative in

any

sense.

Although wakaru has been treated as a stative predicate,

it

has been a problem for many analyses because it can occur in an imperative mood or with V-oo-to-suru, ('try to V'), which should be incompatible with stative predicates: (30) a. Boku no kimoti o/*ga wakare (to wa iwanai). I POSS feelingsACC/NOMunderstand (not say) ' (I'm not telling you to) understand my feelings.' b. Kimi wa boku no kimoti o/*ga wakar-oo-to-si-nai. you TOP I POSS feelings ACC/NOM understand-try-not 'You don't try to understand my feelings.' Note that

in (30)

the nominative

marking in place

of the

accusative marking is not possible. These

facts about

straightforwardly

if

wakaru can we

postulate

wakaru's one [+stative] and wakaru,

like

any other

be

accounted for that

there

rather are

the other [-stative].

simple stative

predicates,

two

Stative takes

dative or nominative for the first NP and nominative for the second NP.

The non-stative one, like any other non-stative

predicate,

takes nominative for the first NP and accusative

.165 for the

second NP,

and can

occur with imperative

mood or

-oo-to-suru. 5. Conditions for the reanalysis Of the putative cases involving the type of reanalysis we are postulating for stative predicates,

the cases with the

predicates suffixed with -tai seem to be most actively used, We

will see

alternating

below case

that a marking

conditions for reanalysis. conditions exist

number can

of be

conditions for best

understood

the as

It is hard to show that parallel

for -reru also.

It is

conceivable that

this is due to the fact that

a sentence with -reru can have

an ergative

well as accusative

case pattern as (See

nominative ones. will

only note

factors

do

in this

affect

Section 6 for the section for

whether

the

and double

discussion.)

V-reru that

reanalyzed

We

semantic

structure

is

preferred or not. 5.1. Distance between the second NP and the predicate There is a general tendency double

nominative

second

NP is

whereas

it is

marking

with -tai constructions that

occurs most

placed immediately less acceptable

naturally

preceding the if there

is a

if

the

predicate, constituent

intervening between the second NP and the predicate. is illustrated by the following examples: (31) a. Boku wa nekkorogatte terebi ga mi-tai. I TOP lying down TV NOM watch-want 'I want to watch TV, lying down.' b.??Boku wa terebi ga nekkorogatte mi-tai.

This

166

c. Boku wa terebi o nekkorogatte mi-tai. The

contrast between

Since

(3la-c)

the reanalysis

we

follows

from our

are stipulating

is

proposal.

done on

the

surface constituent of [NP V]-tai, in order for the NP to be interpreted [V-tai],

as the the

predicate.

argument

NP must In other

be

of

the reanalyzed

interpreted

words,

the

predicate

as sister

reanalysis

to

is a

the local

process, and we will see more evidence for this point below. A similar

situation holds with

the sentences where

both a

direct object and an indirect object are present: (32) a. Boku wa Hanako ni purezento o age-tai. I TOP Hanako DAT presentACC give-want 'I want to give a present to Hanako.' b. Boku wa purezento o Hanako ni age-tai. c. Boku wa Hanako ni purezento ga age-tai. d.?Boku wa purezento ga Hanako ni age-tai. Again, of

it is rather awkward for the dative NP,

the verb

alone and

not V + suffix,

an argument

to separate

the

nominative NP and the predicate, as shown in (32d). This

fact about

the

distance between

the

NP and

predicate has been noted by others (Shibatani 1979, 1971, etc.).

the

Yoshida

When we look at more data, however, we notice

that it is not only the distance that actually matters. the following example: (33) Boku wa terebi ga totemo mi-tai. I TOP TV NOM very see-want 'I want to watch TV very much.'

See

167 The crucial difference between (3lb) that the intervening modifying the verb totemo in

(33)

constituent in (3lb) miru only,

modifies

adverbial in (33)

and (33)

while the

mitai as

is consistent

is the fact

is

an adverbial

degree expression

a whole.

Hence,

the

with the interpretation of

the suffixed

predicate as a whole

sister to it.

The adverbial in

and the preceding (3lb),

NP as

on the other hand,

contributes to separate the verb and the suffix by modifying just the verb. down to

Thus the distance condition

the following:

the predicate

seems to boil

and the

second NP

cannot be separated by an argument or a modifier of the verb alone.

In other words, the reanalysis can involve only the

predicate and

the second

elements can occur

NP.

The

fact that

those other

preceding the second NP even

when it is

nominative shows the locality of the reanalysis. What is significant about this condition is that while it can be perceived naturally as a condition on the reanalysis, it cannot be easily accounted the previous predicate, predicate as

analysis with the

for otherwise; the feature

feature percolation

a whole cannot

especially in

[+stative] on

of the

suffix to

be conditioned by

the the

the surface

word order. 5.2. Conjunction and comparatives we have mentioned above that -tai can attach to conjoined V' 's (cf. (18) and below).

If our hypothesis is correct, it

168 predicts

that conjoined

marked with

V''s

cannot

a nominative case,

have the

since [NP

second

ga V] is

not a On the

constituent under the reanalyzed surface structure. other

hand,

an

accusative-marked

oblique-marked NP)

and

a verb is a

NP

(or

NP

any

other

V' constituent.

This

prediction is borne out. (34) a.*Boku wa [ koocha ga nomi, keeki ga tabe]-tai. I TOP tea NOM drink cake NOM eat want 'I want to drink tea and eat cake.' b. Baku wa

koocha o nomi, keeki o tabe]-tai.

c. Boku wa koocha ga nomi-tai-shi, keeki ga tabe-tai. Note that this

fact also cannot be accounted

feature on the predicate.

for using the

By the time conjunction reduction

applies,

the case marking on the second NP has already been

changed.

As shown by (34c), however, nominative marking on

the second NP is grammatical if the conjunction includes the suffix as

well~

generating

(34b),

problem would conjunction of assuming

so there is no way of blocking (34a), while in

arise in

the previous

analysis.

an approach

(such as

deep structure

-tai subcategorizes

[+stative],

which ,

be no non-ad

and

(by metarule),

hoc way of preventing

where

postulated~

bears a

via percolation to V,

nominative marking on the NP

GPSG)

constituents is for V'

A similar

feature

can trigger a there seems to

the feature percolation

in case V' consists of conjoined V''s as in (34b). The distance condition discussed in the preceding section allows the latter

object NP of (34b)

to

be interpreted as

169 the

argument of

acceptable~

[V-tai],

and

the

following sentence

is

marginal to some speakers, but definitely better

than (34a): (34) d. Boku wa koocha o nomi, keeki ga tabe-tai. This,

again,

shows that the reanalysis of the predicate is

local and not across-the-board. Similarly,

comparative

constructions

such

as

the

following support our proposal: (35) a. Boku, koohii yori biiru ga nomi-tai. I cofee than beer NOM drink-want 'I want to drink beer rather than cofee.' b. Boku, biiru ga koohii yori nomi-tai. (36) a. Boku, koohii yori biiru o nomi-tai. b.*?Boku, biiru o koohii yori nomi-tai. To my knowledge there has been

no explicit analysis of such It is nevertheless

comparative constructions in Japanese. plausible to

think that

(35a)/(36a)

are

more basic

(35b)/(36b) respectively,• and that (35a)/(36a) to

the

syntactic

structures formation

(37a)/(37b) (or

a

below

semantic

than

are related

respectively interpretation)

by

a of

comparatives:

•one piece of evidence comes from the fact that the NP marked with yori can precede the other NP in all contexts, but the other order is not always possible. (i) Boku wa Tokyo yori Kobe ni sumi-tai. I TOP than LOC live-want 'I want to live in Kobe rather than in Tokyo.' (ii)*? Boku wa Kobe ni Tokyo (ni) yori sumi-tai. It seems that NP-yori can be postposed only when the other NP's case is either nominative or accusative (cf. the following footnote).

170 ( 3 7) C\_,

s

N~'

~

\3c1::.LL

te>-..

'-1'

~v

NP

~

V">>Y'I

1,:..,.,1,;:, '-)v'r"i llJ,> ru.. o

And finally,

the rule responsible for the constituent order

of (35b)

(36b)

a~d

can be

comparative phrase NP-yori

postulated as a movement of the to the position next

to V v.•hich

is sister to the other NP under comparison:'

'There is no space to discuss the validity of the analysis as shown in (36), but let me mention a few points of relevance. First, Lhe stipulation NP-yori NP-(£E) is a co~stituent while NP-(g§_) N?-yori is not -is supported by the fact that the former can be the asserted element in a cleft construction (which can contain only one NP), while the lat:er cannot. (i) Boku ga romi-tai no wa koohii yori biiru da. I NOM drink-wantCOMPTOP cof. than beer COP 'What ! want to drink is beer rather than coffee.' (iil *Baku ga nomi-tai no wa biiru ga/o koohii yori da. Secondly, this comparat ve phrase attachment to V' bears some similarity to so-ca led 'quantifier floating' in the sense that a d~gree/quant ty expression (adverbial) attaches

171 ( 38)

s

0-.

b.

/---_ NP

'V' ~

~o'f.u.

NP 'o",·,.,..""~C)..

V' ~

Actv.

V

1::. co'.-;,-, "\o'~'; \'\0"'·,.'\o.\

Now we can pinpoint what is ( 38b),

rhe

V-tai,

unlike

se~antic

moved NP-yori the case in

scope.'

discussion,

modifies only (36a)/(38a),

To summarize this

we can say that

(35b) and (36b)

wrong with (36b).

As shown in

the verb giving

rather complex

and not the wrong line of

the crucial difference between

is that the scope of the comparative phrase

to the V', and in both cases the movement is largely restricted to the cases when che 'host' NP of the adverbial {before it moves) bears either nominative or accusative case. And finally, the stipulation that the moved NP-yori is a modifier to V is consistent with the use of yori as a co~parative prefix to various predicates, which is most likely to be due to a reanalysis of this particle from a postposition to a prefix as illustrated below: [[x yori] ookii] ---> yori-ookii 'bigger' where x is arbitrary or understood. 'Contrast (36b) with the following, where the adverb takusan 'much' makes the verb appropriate for comparative construction, and hence the sentence with the same word order becomes acceptable. (i) Boku wa biiru o koohii yori takusan nomi-tai. 'I want to drink greater amount of beer than coffee.' '-.}.

v ·.--- "-....

~. N?

lo..·,

'V

o·,\ru..O 0..