DOCUMENT RESUME
ED 056 557 FL 002 467
AUTHOR Oh, Choon-Kyu TITLE Aspects of Korean Syntax:Quantification, Relativization, Topicalization,and Negation. INSTITUTION Hawaii Univ., Honolulu. Dept.of Linguistics. PUB DATF Aug 71 NOTE 235p.; Doctoral dissertationpublished in Working Papers in Linauistics,v3 n3 Jun 1971
EDRS PRICE MF-$0.65 HC-$9.87 DESCRIPTORS *Connected Discourse; *DeepStructure; Doctoral Theses; Grammar; *Korean;Linguistic Theory; Negative Forms (Language); *Semantics;sentences; Surface Structure; Syntax;Transformation GenerativQ Grammar; Transformations (Language);*Transformation Theory (Language)
ABSTRACT By offering solutions tolong-standing problems like quantification, relativization,topicalization and negationin Korean syntax, the presentdissertation aims to show thelimitations of any approach which concentrateson the sentence as alinguistic unit or which takes semanticsfo be interp-etative. Onepossible solution suggested here is atopic-by-topi.J approach, with theidea that there are two basic typesof sentences: introductorysentences which introduce the existence of anobject or fact, and other sentences which assertthings other than theexistence of that item. Chapter two covers quantification;chapter three discusses restrictive and non-rertrictiveclauses; chapter four considers topicalization; and chapter fivedeals with negation. Appendix one discusses formal constraints onthe deletion of reflexive pronouns and considers the roleof presupposition in grammarin the area of pronominalization by deletion. Otherappendixes list the rules used in the paper and Koreanparticles. A bibliography isincluded. (Author/W) From: Ubiking Papers in Linguistics;Vol. 3, NO. 32 June 1971.
ASPECTS OF KOREAN SYNTAX: QUANTIFICATION, RELATIVIZATION,
TOPICALIZATION, AND NEGATION
A DISSERTATION SUBMITTED TO THE GRADUATEDIVISION OF THE UNIVERSITY OF HAWAII IN PARTIALFULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
IN LINGUISTICS
AUGUST 1971
U.S. DEPARTMENT OF HEALTH, EDUCATION & WELFARE
OFFICE OF EDUCATION
THIS DOCUMENT HAS BEEN REPRODUCED EXACTLY AS RECEIVED FROM THE PERSON OR ORGANIZATION ORIGINATING IT.POINTS OF VIEW OR OPINIONS
STATED DO NOT NECESSARILY REPRESENT OFFICIAL OFFICE OF EDUCATION POSITION OR POLICY.
BY
Choon-Kyu Oh
Dissertation Committee:
Charles-James N. Bailey, Chairman Lewis S. Josephs
"' Gary J. Parker
014 0 Copyright (4,1i; 1971
by Choon Kyu Oh
"PERMISSION TO REPRODUCE THIS COPYRIGHTED MATERIAL HiS BEEN GRANTED
BY(-110n K1u.0V\
TO ERIC AND ORGANIZATIONS OPERATING UNDER AGREEMENTS WITH THE U.S. OFFICE OF EDUCATION. FURTHER REPRODUCTION OUTSIDE THE ERIC SYSTEM REQUIRES PERMISSIONOF ME COPYRIGHT OWNER." DEDICATION
To my teacher Charles-James Nice Bailey, whohas been probably the firstthesis chairman with so much interestand care in his student andwho has awoken in me an intensive interestin the search of the principles of language, and
to my wife Soung-Ei Oh, who has willinglyand lovingly taken the severe ordeal of four years, sacrificingalmost everything thatbelongs to her, and even financially supporting mystudy with the moneyexchanged with painful labor,
I dedicate this thesiswith deep appreciation. FOREWORD
One of my most pleasantexperiences as a faculty member and as a co-editor of these Working Papers has beenthe supervising of Choon-Kyu
Oh's doctoral dissertation. This studept's innate knack forhandling the complex materials of moderngenerative semantics has enabledhim in a brief eight months of concentratedlabor to produce a work that will make many contributions to thetheory of generative semantics.
I have been mainly concernedwith putting all relevant materialsinto
Oh's hands and with questioning his proposalsin any way in which I have thought they were weak,since I have had to leave thetechnical- ities of Korean syntax to my colleagueLewis Josephs. Both of us, together with Gary Parker, the otherco-editor of this publication, have gone over every word in this writingwith as critical an eye as possible. The result is one that we feel hasfew weakneses and many
strengths. We are pleased to commend it to thereaders of this series with the certainty that they will find manythings in it to interest
and profit them. We sincerely hope that Ohwill reap the rewards of
his diligence and remain true to the qualityof work he has exhibited
here.
C.-J. N. Bailey ABSTRACT
Chomsky used to claim thatonly the syntactic componentis generative, while the phonologicaland semantic components are
supposed to be solely interpretative. The base eomponent was responsible for meaning, andtransformations were not supposedto
change meaning. As the discipline oflinguistics has advanced
since then, we have come to seethe limitations of thisapproach,
and linguists such asLakoff, Ross, and McCawleyhave claimed
that the semantic componentis generative. Since the adoption of
generative theory in syntactic analysis,almost all linguistic
arguments have centeredaround the sentence. However, this
sentential approach has resultedin the insolubility ofmany'
problems, such aspronominalization, topicalization,quantification,
relativization, and negation.
By offering bettersolutions to long-standingproblems
and like quantification,relativization, topicalization,
negation in Korean syntax,the present dissertationaims to show
the limitations of anyapproach which concentrates onthe sentence
as a linguisticunit or which takes semanticsto be interpretative.
It is true that at thepresent stage there are still manyproblems
to be solved within thegenerative-semantic, discourse-centered
approach (the problem ofdeciding the boundary of adiscourse being
only one of them). One possible solutionsuggested here is a
topic-by-topic approach--viz, theview that there are twobasic
types of sentences: an introductorysentence which introduces the existence of an object orfact, and other sentenceswhich assert things other than theexistence of that item.The same sentence non-introductory for another can beintroductory fer one NP and is NP in the largersentence. Nat:=ally, an introductory sentence ordered before the othertypes of sentence.
In Chapter Two, Carden'sanalysis of quantificationis It is reviewed and itslimitations are indicated andcircumvented. shown how the concept ofthe introductory sentenceenables the problems related to grammar correctlyto predict and explain many suggested approach is tested quantification. The validity of the contain quantifiers on negating andinterrogating sentences which adjective plus.a and other indefinitePhrases consisting of an peculiar properties of noun. Carden's explanations of the
quantifiers observed byhim are carried furtherby showing that
quantifiers cannot bepredicates of introductorysentences.
Adjectives which occur in anon-introductory sentence arenegated
exactly the way adjectives arewhen moved into anintroductory
sentence by a rule thatgenerates non-restrictiverelative clauses, In Chapter which rule is alsoresponsible for movingquantifiers.
Three, restrictive ananon-restrictive relativizaticns are structures are suggestedfor them discussed. Different underlying is shown to and juztified. RESTRTCTIVE-RELATIVE-CLAUSE FORMATION sentence into the be a copying rule. It copies an introductory opposite of the rule, following sentence. This copying rule is the
NON-RESTRICTIVE-RELATIVE-CLAUSE FORMATION, exceptthat the former is not a simple pc-mutation rule.
The definition of definitenessadopted in this dissertation suggests a neat solution to an age-oldproblem in Korean--nn and ka as subject particles(the counterparts of Japanese wa and22), which is discussed in Chapter Four. In Chapter Five, the author shows how a NEG-INCORPORATION ruleis needc,C( for the correct descx.iption of negation in Korean. In an Appendix, the role of presupposition in grammar isillustrated with one of the major processes in Korean syntax,pronominalization by deletion.
The writer tries to show throughouthow the generative- semantic approach alone can handlein a natural manner the problems discussed here, by importing the notionof presupposition into syntax. LIST OF ABBREVIATIONS*
COM COMMATIVE
DIR DIRECTIONAL
DLM DELIMITER
NEG NEGATIVE MARKER
NOM NOMINALIZER
NT NOT-TRANSPORTATION
0 OBJECT-MARKER
PL PLURALIZER
QUANTIFIER
0T QUOTATIONAL
QUES QUESTION
REL RELATIVIZER
RET RETROSPECTIVE
SUBJECT-MARKE71
TOPIC-MARKER
*For the names ofrules, see Appendix II. TABLE OF CONTENTS
ABSTRACT
LIST OF ABBREVIATIONS vi
Chapter I INTRODUCTION 1
Chapter II QUANTIFICATION 4
2.1 Introduction 4
2.11NOT-TRANSPORTATION 4
2.12Verb of denial in Korean and English 10
2.13 EQUI-NP DELETION and REFLEXIVIZATION 13
2.14CONJUNCTION REDUCTION 20
2.2 The derivation of indefinite NP's modified by
adjectives 31
2.21 Other negations and questions 40
2.22EQUI-NP DELETION and REFLEXIVIZATION 49
2e23 CONJUNCTION REDUCTION 56
Chapter III RELATIVIZATION 60
3.1 Introduction 60
3.11 Introductory E,entence and Lelated matters 62
3.12RESTRICTIVE CLAUSE FORMATION and the definiteness
of NP's 66
3.13NR-FORMATION 87
3.2 Some possible objections 95
3.21 Direction ofeMbedding in NR-FORM and the problem,
of ambiguity of a NR 96
3.22Stacked restrictive relative clauses 103 110 3.3 Other related problems
3.31 Derivation of generic nouns
3.32 Identity and referentiality of NP s 119
3.33 Implications made by the hierarchlof anaphora 124
Chapter IV TOPICALIZATION 127
4.1 Introduction 127
4.11 Definite NP and TOPICALIZATION 129
4.12 Further restriction on TOPTCALIZAT )14 134
4.13 TOPICALIZATION as a postcyclic rul 139
4.14 Restriction on the applicability o,
TOP1CALIZATION 144
4.2 The so-called objective ka (or se 148
Chapter V NEGATION 159
5.1 Introduction 159
5.2 The NEG-TRANSPORTATION approach . 161
5.21 The illegitimacy of NT with hats 162
5.22 The theoretical implications of ha. 7LETION 169
5.3 The NEG-INCORPORATION approach 173
5.31 The correct derivation of negation 177
5.32 Justification of 4-ht )EG-INCORPORAT ON 178
APPENDIX I FORMAL CONSTRAINTS ON THE DELETION OF REFLEXIVE
PRONOUNS 185
APPENDIX II LIST OF RULES 197
APPENDIX III LIST OF KOREAN PARTICLES . 199
NOTES 201 BIBLIOGRAPHY' 10 218 CHAPTER ONE: INTRODUCTION
the writer has kept In writing thepresent dissertation, limitations of the following goalsin mind: to show some of the of a language and concentrating on the sentencein the description of sentences, aswell as of ekcluding semanticsfrom the generation in Korean can behandled with to show how somelong-standing problems The writer mostlyfollows the new insightsin a non-ad-hoc manner. linguists like Lakoff, abstract approach tosyntactic analysis of particularly in the Ross, and McCawley,with slight modifications, information in formulating manner ofutilizing discourse presuppositional informationin transformational rules: I have used transformational rules, aswell as in the structuraldescription of Recently, the advantagesof the conditions ontheir applicability. grammarians like Longacreand discourse analysishave been noticed by not taken fulladvantage his followers, butthey have unfortunately available by generative- of the discoveriesand insights made
transformational theory. writer has To achieve thegoals mentioned atthe outset, the problems in language chosen to discussthe most controversial relativization, which are descriptiontoday--quantification and respectively--employing both dealt with in ChaptersTwo and Three, arguments. It kaean and English data to illustrateand corroborate processes asquantification is the author'sbelief that such deep languages throughoutthe and relativizationare shared among this writing a,--e world. But because of this,the rules in formulated in such a way that thesimilarities between the English and Korean formulations standout conspic-,hously. Explicit characterizations of such terms asdefiniteness, specificity, and correferentiality have been necessitatedin order to account for various aspects of quantificationand relativization which havebeen obscured by the loose use of suchterms in the past.
I have introduced a newconcept, the introductorl sentence
(abbr. intro-S), which is eitherasserted or presupposed in a discourse. This has been done with a view toclarifying the definiteness of NP1s. An intro-S is a sentencewhich establishes
or registers theknowledge of the existence of anyfact or thing
denoted by a certain NP.Once this knowledge has beenregistered
in the mind of a hearer, that NPbecomes definite. If the knowledge then such that a speaker possesses hasnot been shared with a hearer,
an NP is merelyspecific, not definite. NP's have to be at least
specific71 and also identically indexed to becorreferential, as
will be shown.
Many linguists have noticedthe limitations of current thecry,
particularly in its distinctionbetween optional and obligatoryrules.
There has been no formalmechanism to indicate thedegree of
optionality of a rule.The author has not supplied anysuch
mechanism in the present work,but has at least attemptedto show
some of thecharacteristics of optionality.Very few rules are
optional in the sense of beingin alsolutely free variation. Most
rules which have been knownto be optional are infact rules 3 concerning which speakers do not sharethe presupposition which conditions the applicability of therule in question (for a concrete example, see Appendix I).
I have been able to explainthe long-debated question regarding the difference between ninand ka (Japanese wa and 22, respectively) by clarifying the feature[definite] and by introducing distinct formal and materialfunctions of NP,s. If
an NP carries newinformation in a sentence, it isused formally;
otherwise, materially,. It is shown in Chapter Fourthat only
materially-used NP,s can be topicalizedby attaching nn. Although
negation in Korean has drawn muchattention from Korean
syntacticians, most of their writingshave been merely descriptive,
and hardly explanatory. The utilization of logicalformulations in
linguistic description, advocated by G.Lakoff and McCawley, has
enabled the author to account for manyfacts concerning Korean
negation that have so far beenneglected.
It is the author's humble hopethat this work will
contribute to the study of syntaxby introducing certainphenomena
from Korean syntax to non-Koreanlinguists that will lead togreater
clarification of the principles ofgenerative grammar and perhapsof
the analysis of English itself. I also hope to make modern
discoveries of generative semantics moreeasily available to Korean
grammarians. 4
CHAPTER TWO: QUANTIFICATION
Introduction behavior 2.1. In his attempt toexplain some idiosyncratic between post- of quantifiers (abbr.Q), Carden (1970) distinguished with other and pre-determiner Q'sand identified post-determiners 1 pre-determiners were higher adjectives. He claimed that the predicates in the deepstructure which are lowered tothe pre-
determiner position by hislowering rule. However, it appears that pre-determiner Q's is the same distinctionmade between post- and
to be made in otheradjectives as well. A review of Carden's
arguments for ahigher-predicate analysis of Q'sin the light of for Q's should not be Korean data revealsthat the distinction made
made in terms oftheir position before orafter the determiner, as
Carden did, but interms of the definitenessof the NP's which they
2 for English, it will be modify. Although Carden's arguments are
shown that the sameprinciple which accounts forEnglish
quantification also accountsfor Korean quantification. In the
sections that f)llow,Korean data will becontrasted with
corresponding Englishdata. Carden (39-40) 2.11. NOT TRANSPORTATION(abbr. NT). negative marker claims that are-determiner Q is negated by a but other (abbr. NEG) attached to anot-transportation (NT) verb, complement clause after a NT kinds of adjectiveswhich occlu: in a 3 In other verb are notnega-Eed by such a NEG on theNT-verb. 5 words, la is synonymous with lb, butnot with lc; and 2a with 2b, but not with 2c; ex. 3a, however, is synonymousnot with 3b, but with 3c:
(1) a. John doesn't think the attractivegirls left.
b. John thinks the attractivegirls 212221 (= did not leave).
c. John thinks theunattractive girls left (notattractive).
(2) a. John doesn't think the many girlsleft.
b. John thinks the many girls stayed.
c. John thinks thefew girls left.
(3) a. John doesn't think that many (ofthe) girls left.
b. Jbhn thinks that many girlsstayed.
c. John thinks thatnot many girls left.
Carden (40-41) further arguesthat the Q many which is negatedin 3c
must come from a higher predicatein the deep structure, since the
NEG attached to a NT-verbnegates the highest predicatein the
complement sentence, as shown here:
(4) a. John doesn't think that theboys who left will catch the
train.
b. John thinks that the boys wholeft will not catch the
train.
c. John thinks thatthe boys who staled will catchthe
train.
The correctness of the abovegeneralization is clearly demonstrated
by the fact that 4a is synonymouswith 4b, but never with 4c.
However, in order toascertain the correctness of thedistinction
5 6 boLwoun a pre-determinor Q, on the one hand, and a post-determiner Q and all other adjectives, whether definite and indefinite, onthe other hand, let us investigate some parallel cases in Korean.
In Korean, the scope of a NEG depends on the presence or absence of the particle nin 'delimiter' (abbr. DLM) after the
4 nominalizer. If the particle is present, the scope of a NEG includes the entire complement; otherwise, only the highest
predicate of the complement sentence is negated. (Full discussion
of the various kinds of negation is postponed until ChapterFive.)
Only two types are considered for present purposes. Thus, the
following two sentences--the first with nIn and the second without
it--are not synonymous, as indicated by the English translations:
(5) John-ka ka-ci ani-hayessta
Jchn-S go-NOM not-did
'John did not go (= 'John stayed').
(6) John-ka ka-ci-nln ani-hayessta
John-S go-NOM-DLM not-did
'It is not the case that John went.'
Now, consider the following:
(7) a. John-ka ka-n-kes-nth ani-ita
John-Sgo-REL-thing-DLM not-is
lIt was not John who went.'
b. John-ka ka-n-kes-nin ani-ita
'John did not go.' (8) a. kyeyppl-n yeca-ka nathana-n-kes-nn ani-ita
the pretty-REL girl-S appeared-REL-thing-DLMnot-is
'It was not the pretty girl who appeared.'
b. kyeypiA-n yeca-ka nathana-n-kes-nin ani-ita
'The pretty girl did not appear.'
(9) a. yeypp&-n yeca-ka nathana-n-kes-nin ani-ita
'It was not pretty girl(s) that appeared.'
b. *yeyppl-n yeca-kangthana-n-kes-nln ani-ita
?'Prety girls did not appear.'
(10) a. mSnhi-n salam-ka ka-n-kes-n&n ani-ita
Many-REL people-S went-REL-DLMnot-is
'It was not many people that went.'
b. *manhk.-n salam-ka ka-n-kes-ninani-ita
7,Many people did not go.'
In 7 and 8, the embeddedsubjects are definite, whereas thoseof 9 and 10 are indefinite.This definite-vs.-indefinitedistinction in the subject NP's of embeddedsentences seems to be responsiblefor the difference in the readingsof 7 and 81 on the one hand, and9 and 102 on the other hand, that is, with the former, the NEG outside the complement sentence cannegate either the sub"ect or the predicate of the embedded sentence,depending on where the heaviest
accent is. Only the accented part is negated.However, only one
interpretation is possible for anindefinite subject NP of an
embedded sentence, viz. NEG alwaysnegates the sUbject NP.
For additional support of theabove analysis,--Iet us
consider one other kind of negation: 17 (11) a.Jci;hn-ka kassta-ko-nin sayngkakha-ci ani-hanta
John-Swent-QT-DLM think-NOM not-do
'I do not think that it was John that went.'
b. John-ka kassta-ko-nln sayngkakha-ci ani-hanta
'I think that John did not go.'
(12) a. kg yeyppi-n yeca-t11-ka nathanassta-ko-nin sayngkakha-ci
the pretty-REL girl-PL-S appeared-QT-DLM think-NOM
ani-hanta
not-do
'I do not think that it was the pretty girls that
appeared.'
b. ki yeyppi-n yeca-t11-ka nathanassta-ko-nln sayngkakha-ci
ani-hanta
'I think that the pretty girls did not appear.'
(13) a. mgnhl-n salam-til-ka kassta-ko-nln sayngkakha-ci
many-REL man-PL-S went-QT-DLM think-NOM
ani-hanta
not-do
'I do not think that it was many men that went.'
b. *manhi-n salam-til-kakAsta-ko-nin sayngkakha-ci
ani-hanta
'I think that many men did not go.'
(14) a. yeypp&-n veca-Ul-ka nathanassta-ko-ninsayngkakha-ci
ani-hanta -- II do not think that it was prettygirls_that appeared.' 8 9
(14) b. *ycypiA-n yeca-tl-kan'thanassta-ko-nn sayngkakha-ci
ani-hanta
'I think that pretty girlsdid not appear.'
Again, in 11 and 12, which havesurface sentence subjects which are definite, the most heavily accentedwords are negated--either the subject or the predicate of thecomplement sentence.But in 13 and
14, which have indefinitesubjects, the predicate ofthe complement
(viz. kassta or nathanassta) cannotbe negated with a NEG attached to an NT-verb of the matrixclause.Thus, both the (a) and(b)
sentences of 11 and 12 aregrammatical, whereas the (b) sentences
of 13 and 14 are not.
What is important here isthat a distinction is notdrawn
between pre-determiner Q andthe rest but between definiteand
indefinite subjects, regardless ofwhether they are modified by Q
or some other kindof modifier. Carden's higher-predicateanalysis
seems to beapplicable not only to 10 and 13,but also to 9 and 14,
where a non-Q adjectivemodifies an indefinite subject.
Does this insight from Koreansyntax apply also to English?
The question seems to be areasonably simple one, if weconsider 15:
(15) a. John doesn't think thatattractive girls left.
b. John thinks thatattractive girlsstanch
c. John thinksthat non-attractive girlsleft.
For 15a, the interpretation seemsto proceed in thefollowing
manner: the universe of girls isdivided into two groups;
attractive ones and non-attractive ones. If the embedded subject 1 9 10 were definite, as in la, ourtopic would be limited to that definite group. All that la says is that none of theattractive group left and we do not commit ourselves toany further information in this sentence. On the other hand, since the embedded subjectis not definite in 15a, this sentence carries--in addition tothe information that none of the attractive group left--the sensethat some girl(s) left and that girl(or those girls) must therefore belong to the non-attractive group.In other words, 15a is
synonymous with 15c--not with15b, as would be expected, if Carden's
analysis were correct. In short, if the higher-predicateanalysis
applies to any pre-determiner GI, it should alsoapply to all
indefinite NP's modified by any adjective.
2.12. Verb of denial in Korean andEnglish.5Carden (41- p. 42) has claimed for 16 and 17, but not 18,that we have to
presuppose the existenceof Whig candidates and many candidates,
respectively, but deny that they won.
(16) John denies that the Whig candidates won.
(17) John denies that the many candidates won.
(18) John denies that many candidates won.
On the other hand, 18 presupposesthat there are candidates who won,
and denies that there were manyof tnem. At this point, we may
profitably return to further data fromKorean:
(19) a. John-nlnMLN.-ka cukess-im-lil puinhayessta
John-T Mary-Sdied-NOM-0 denied
'John denied that it was Mary thatdied.' 21) 11
(19) b. John-nth Mary-kaciikess-lm-111 puinhayessta
'John denied that Mary died.'
(20) a. John-nth 14 tases/man1-4-n haksayn9-ka ikyess-im-lil
John-T the five/many-REL students-S won-NOM-0
puinhayessta
denied
'John denied that it was the five/many studentsthat
won.'
b. John-nin ki tases/manh-n haksayng-kalicyess-im-111
puinhayessta
'John denied that the five/many studentswon.'
(21) a. John-nin k yere2L-ny122-..-ka nathanass-im-111
John-T the pretty-REL girl-S appeared-NOM-0
puinhayessta
denied
'John denied that it was the p7ettygirl that appeared.'
b. John-nln kl yeyppn yeca-kangthanass-im-lil
puinhayessta
'John denied that the pretty girl appeared.'
(22) a. John-nZlnmgnhi-n/tases haksayng-ka cukess-lm-lil
John-T many-REL/five students-S died-MA-0
puinhayessta
denied
'John denied that it was many studentsthat died.'
b. John-nin manhi-n haksayng-kacukess-im-111 puinhayessta MIIIIMENIM111=0 Ijohn denied that many:'students died.' 12
141. puinhayesota (23) a. John-Mtnytlypp4-nyeca-k4 ndanginass-4.6
-0 denied John-IT pretty-REL girl-S appeared-11a
'John denied that it was pretty girls Cat appeared.'
b. John-nin yeypp4-n yeca-ka nathanass-io-141_ puinhayessta
'John denied that pretty girls appeare(
As we see here, numerals seem to behave much like c,s, and so do indefinite NPIls modified by descriptive adjectives. In 19-21, where the underlined NP's are definite, we have amblguit as to which part of the embedded sentence is negated. This ambigultf again willbe removed by auxiliary devices, viz. proeodic featurt 3. In 22 and 23, however, only the adjective or Q modifying the sajectNP is negatable; note that /2522i= 'pretty' in 23 is rJt even aQ. The special treatment of 416 as opposed to descriptiie adjectives, seems to lose ground at this point.This is not eAe to a syntactic phenomenon peculiar to Korean, 411 is st,',n by the following parallel English sentences:
(24) John denies that rich candidates won.
(25) Jdhn denies that the rich candidates won.
Although the speaker assumes in 25 that there wererich candidates and denies their having won, in 24 he denies thatthe assumed victorious candidates were rich. In the latter case, rich behaves
just like Carden's pre-determiner Q. If it can be justified that maxn in 18 is derived from thepredicate of an underlying higher
sentence, the same process may legitimately be inferredfor the
derivation of 24. 13
2.13. EQUI-NP.DELETION and REFLEXIVIZATION. Carden (45) makes the following peculiar claim concerning26:
What is importart to the present argument is that, insentences like (C260), where the identical,co-referential noun phrases are modified by a trueadjective ("Whig"), the following facts hold: A. The sentence without deletion([26a]) is ungrammatical. B. The sentence without deletion([26a1) is synonymous with the sentence withdeletion ([26b]).
Carden claims further that the same is trueof only a post-determiner
Q (as in 27), but that neither 'fact Al nor'fact Bs is true of a pre-determiner'Q (as in 28):
(26) a. *The Whig candidatesi expect theWhig candidatesi to be
elected.
b. The Whig candidates expect to beelected.
(27) a. *The many candidatesi expectthe many candidatesi to be
elected.
b. The many candidates expect to beelected. Ma.Allthecandidates.expectallthecandidates.to be
elected.
b. All the candidates expect to beelected.
Now, let us turn to Korean data:
(29)a.mo-U-nkyoin-t&l-nki.moti-nkyoin-t11-ka1 chenkuk-ey
all-REL believer-PL-T all-REL believer-PL-Sheaven-to
ka-l-kes-111 kitayhanta
go-REL-thing-0 expect
'All believers expect all.believers to goto heaven.'
2:3 14
(29) b. mot-n kyoin-.U1-n&n (caki-ka) chenkuk-ey ka-l-kes-lil
kitayhanta
'All believers expect to go to heaven.'
A(30) a. *John-nIni John-kai chenkuk-ey ka-l-kes-111 kitayhanta
1Johni expects Johni to go to heaven.'
b. John-nln (caki-ka) chenkuk-ey ka-l-kes-lll kitayhanta
'John expects to go to heaven.'
In Korean, the deletion of any embedded NPcorreferential with the matkix NP seems to be semantactically conditionedin most cases.
This matter is fully discussed in the next chapter.But, for the present purpose, I will consider only instances inwhidh the deleted
NP or reflexive pronoun is correferentialwith the matrix NP.Thus,
even though 30a is ungrammaticalor at least very unnatural, a 4
forced interpretation could make it synonymous with30b. On the
other hand, 29a is not only grammatical,but semantically different
from 29b. This seems to correspond well with the abovegeneralization
of Carden's.However, one very interesting phenomenonis revealed
in the use of the reflexive pronoun is29b: when the NP in the
embedded sentence is correferential with the NPin the matrix
sentence, the reflexive pronoun should besingular, as in 29b. If a
plural form is used, exact correferentialityis lost.
Unlike English, Korean allows what I proposecalling
inclusive as well as exclusive reflexivization. Inclusive
reflexivization is the replacement of a singular NPwith a plural
reflexive pronoun. Exclusive reflexivization is the replacementof 15
These processes are a singularNP with a singularreflexive.pronoun. respectively illustrated in 31and 32:
(31) John-n&n caki-ney-kaikilila-ko Mitko-issta
John-T (him) self-PL-S win-QT believe
'John believes that they(lit. Ithemselves1; i.e.his own
group) will win.'
(32) John-nln cdki-kaikilila-ko mitko-issta
'John believes that hehimself will win.'
and those with him', in 31, caki-ney(or -t1) may mean 'John With the understood as fellag-studentsin his school or thelike.
distinction between inclusiveand exclusivereflexivization exists only in 29, clarified, we can.easilyexplain why ambiguity varieties of but not in 30.For it is only in29a that the two
reflexivization are possible. In other words,beth types of
reflexivization are allowedonly where the subjectis not definite,
which fact is madeclear in the followingsentences:
(33) a. Samoa-salam-nkiSamoa-salam-kamoyoktangha-nin-kes-111
Samoa-people-TSamoa-people-Sbe-ridiculed-REL-thing-O
mos-chamninta
cannot-stand
'Samoans cannot bear tohave Samoans ridiculed.'
b. Samoa-salam-nincaki-ney-kamoyoktangha-nin-kes-111
mos-chamninta
'Samoans cannot bear tohave their fellowpeople
ridiculed.' 25 16
(33) c. Samoa-r;alam-nln (caki-ka) moyoktangha-nin-kes-lga
mos-chamninta
'Samoans cannot bear to have themselves ridiculed.'
(34) a. yongkamha-n salam-nin yongkamha-n salam-lilchanmihanta
brave-REL people-T brave-REL people-0 admire
'Brave people admire brave people.'
b. .yongkamha-n salam-nin caki-ney (or-t1l)-lll
brave-REL people-T self-PL-0
chanmihanta
admire
c. yongkamha-n salam-nincaki-lll chanmihanta
'Brave people admire themselves (= 'Everybrave man
admires himself') .
When an indefinite member of a presupposedgrouping is chosen as the subject of the higher sentence, acorreferential
subject of the embedded sentence may bereflexivized exclusively
(i.e. in the singular form), as in 33c; if the subjectof an embedded
sentence is correferential with only partof the group denoted by
any NP in the higher sentence,then reflexivization is inclusive,
as illustrated in33b. If a grouping is not presupposed, inclusive
reflexivization is rendered impossible; e.g.34h. But even in such
instances, a non-pronominalized repetition ofthe NP is allowed
only where inclusive reflexivizationwould have been permitted. It
seems that 33a and 34a areparallel to 35 and 361 respectively: 17
(35) a. John-nln. John-ka.Inoyoktangha-nIn-kes-lll mos-chamninta
'John cannot bear to have John.ridiculed., 7
b. John nIn (caki.-ka)moyoktangha-nln-kes-lI1 mos-chamnInta
'John cannot bear to beridiculed.'
(36) a. John-nln. John-lll. chanmihanta 7 john-T John-0 admire
'John. admires John..,
b. John-nln caki-lllchanmihanta
'John admires himself.'
Just as the use of theidentical NP John withoutcorreferentialtty does not justifypronominalization in 36a, so the useof the identical NP yongkamha-nsalam 'brave man' cannotjustify pronominalization in 34a--eft.therin the form of a reflexive pronoun or of completedeletion--unless the :identicalexpression presupposes is some sort ofgrouning, in which caseinclusive reflexivization generalization, allowed. Exs. 33 and 34 also showthat Carden's (45) applicable to a quoted at the beginningof this section, is also
'true' adjective as well as apre-determiner Q4 only if the NPis
indefinite.
To recapitulate, thefactor that determines thedeletability
of repeated NP's (e.g.26-28) and the applicabilityof exclusive
reflexivization (e.g. 29, 30,33, and 34) is not themodifying
function of a 011!, but rather thedefiniteness of the NPmodified.
That this is not peculiarto Korean is evidentfrom the following
English sentences: 27 18
(37) a. Blonde girls .Bxpect blonde girls to bespecially
treated.
b. Blonde girls expect to bespecially treated.
Even though a true adjective occursin 37, neither A or B in
Carden's generalization quoted on page13 holds. In short, Carden
seems to have been wrongin assuming that the two uses ofwhia
candidates and all candidates in 26a and28a are correferential.
They are not.What Carden calls 'fact A' and'fact B1 in the
quotation above both hold true onlywhen the NP modified by an
adjective--Q or not--is definite; ifthe NP is indefinite; neither
claim holds.
With this misconception aboutcorreferentiality corrected,
we ca n. revise thegeneralization concerning sentenceslike 26-28.
The following facts hold ifand only if identical NP'sin a matrix
sentence and in a sentenceembedded in it are correferential:
A. Sentences without deletion areungrammatical. B. Sentences
without deletion are synonymouswith sentences which have
undergone the deletion of the NPin question.Furthermore,
when the NP construction isdefinite, the non.-correferential
NP's refer to two different groupsor to separateindividuals;
whereas when the NPconstruction is indefinite, the non-
correferential NP's refer todifferent members of a group.
Furthermore, thedefinite-indefinite distinctionbecomes
logically impossible withNP's modified by auniversal Q, such as difference between 28a and28b all and every. Thus, the syntactic 19 reflects the followingsemantic difference: 28a can be paraphrased in logical as 38a, and28b as 38b; these inturn may be represented terms as 39a and 39b,respectively:
(38) a. There is a candidatesuch that he wants allthe
candidates to be electedlandthat is true of all the
candidates.
b. There is a candidatesuch that he wants tobe elected,
and that is true ofall the candidates.
(39) a. (,&) ((yy)tx want y beelected] )
b. (yx) (x want1:x be elected]) REFLEXIVIZATION as for Carden (46) madethe same claim for reflexivization are EQUI-NP DELETION. Sentences without within a ungrammatical when identicalcorreferential NP's occur this to be true onlyof NP's single sentence. But Carden claimed post-determiner Q's, as in that are modified bytrue adjectives or
41 and 42:
(41) a. *The foolish menipity the foolish men..
b. The foolish menpity themselves.
(42) a. *The many menipity the many meni.
b. The many menpity themselves.
(43) a. All the menpity all the men..
b. All the men pitythemselves.
(44) a. John pitiesJohn.
b.John.pities himself(= John.).
not correferentialat all, The two occurrencesof men in 43a are 20 just as the two identical NP's in 44a are not.
2.14. CONJUNCTION REDUCTION. So far, I have been treating reasonably clear-cut cases.But the dichotomy between definiteness and indefiniteness of NP's is not sufficient adequately toaccount for the following.An additional feature is required. (45)a.Goodraes.areexplicalarvdgoodrules.are easy to
reed.
b. Good rules are (bobh) explicit and easy toread.
(46)a.Few(ofthe)rules.are explicit, and few (of the)
rules are easy to read.
b. Few (of the) rules are (both) explicitand easy to read.
(47)a.Thefewrules.are explicit; and the few rulesare
easy to read.
b. The few rules are explicit and easy toread.
Carden (48) claims that 45a and 47a, where thenoun-modifiers are
not pre-determiner Q.'s, are amenable toconjunction reduction
without involving any change in the meaning,whereas with 46a, where
a pre-determiner Q isfound, conjunction reduction changes the
sense even when thetwo subjects are correferential.But in what
sense are theycorreferential? Below, I will provide an explicit
definition of correferentiality which will ensurethat non-
correferential subjects will not basubject to conjunction
reduction. The lack of synonymy between 46a and46b, as opposed to
the synonymy of 45a with 45b and47a with 47b, does not seem to
come from the fact that46 alone has a pre-determiner Q, as Carden 21 would like to argue.This is convincinglydemonstrated in 48:
(48) a. All the good rulesi areexplicit, and all the good
rules, are easy to read.
b. All the good ruues are(both) explicit and easy toread. l'aven though thenoun-modifier in question here is apre-determiner change in Ot, conjunction reductionis possible without any apparent meaning.
However, the analyticaltask has not yet been completed.
So far, we have em.ymanaged to explain someproblems in terms of
definiteness and a somewhat clearerunderstanding of
correferentiality, though withoutbeing really explicit. Soma How further questions that shouldbe answered are thefollowing:
does one know whether NP's arecorreferential or not?And what is The the correlation betweencorreferentiality and definiteness?
subjects in 47a and 48a aredefinite and correferential,while
those in 45a and 46b areindefinite and correferential.Clearly,
indefiniteness does not entailnon-correferentiality.Why should than the this be so?This leads us to atwo-dimensional scale rather
single distinction betweendefinite and indefinite NP's. The
additional distinction that crossesthe other seems to be a 6 There are four distinction between 'specificand 'non-specific'.
logical possibilities: r+ definite, + specifiC), E definite,
-specifig. -specifiC1, & definite, + specific], andE- definite,
Definiteness in the sense usedhere signifies what isalready
registered in theknowledge of both the speaker andthe hearer. 22
Specificity signifies something which, although not necessarily
wiyamM11 registered in the knowledge of the hearer, nevertheless has a determined identification. This definition logically obviates the possibility of the combination r, definite, pecific] from ever occurring. In other words, definiteness implies specificity, although the converse is not necessarily true. Proper nouns, for example, are definite, and therefore necessarily specific.
Now, we can define correferentiality as the relation between
NP's which are identically indexed and are both& specific), regardless of whether they are definite or not; e.g. the subjects of
47a and 48a, which are both definite, are correferential. But NP's do not have to be definite to be correferential, as is shown by the
subjects of 45b and 46b--which are u specific] but not C+ definite].
The subjects of 46a, however, are not even[+ specific] and
therefore are not correferential; consequently, they cannot be
definite.Let us turn to some relevant Korean data:
(49) a. yek-ey-to manhi-n salam-ka iss-ko, keli-eyse-tomanhl-n
station-at-toc, many-REL people-S are-and street-on-too
salam-ka poiessta
many-REL people-S were-seen
'Many people were at the station, and many people were
seen on the street.'
b. manhl-n salam-ka yek-ey-to iss-ko keli-eyse-topoiessta
'Many people were at the station and also were seen on
the street.' 23
(50) a. ecey-to manhi-n salam-ka oass-ko, neyil-to manh-n
yesterday-too many-REL people-S came-and tOmorrow-too
salam-ka o-lkesita
many-REL people-S will-come
'Many people came yesterday, and many people will come
tomorrow.'
b. manh&-n salam-ka ecey-to o-ass-ko neyil-to o-lkesita
many-REL people-S yesterday-too came-and tomorrow-too
will-come
(51) a. hakkyo-ey-to mikuk-salam-ka iss-ko, samuso-ey-to
school-at-too American-people-S are-and office-at-too
mikuksalam-ka pointa
Americans-S are-seen
'Americans are at the school, and Americans are to be
seen at the office.'
b. mikuk-salam-ka hakkyo-ey-to iss-ko samuso-eyse-to
American-people-S school-at-too are-and office-at-too
pointa
are-seen
(52) a. ecey-to John-ka oass-ko, neyil-to john-ka o-lkesita
Yeste'cday-too john-S came-and tomorraw-too john-Swill-come
'John came yesterday, and John will cometomorrow.'
b. John-ka ecey-to oassko neyil-to o lkesita
John-S yesterday-too came-and tomorrow-toowill-come
Each of the pairs of sentences in 49-52is synonymous, despite 24
- differences in definiteness and also in specificity. These,
however, should not be considered to be counterexampaesto what has
been claimed above. For in Korean, as long as some contrasteditems
(e.g. the Eb3 sentences of 49-52) are overtly presentin the
conjuncts, correferentiality is not required forfurther reduction
(viz, reduction by CONJUNCTION REDUCTION) to occur. But when such
items are not present, correferentiality seems tobe the
determining factor in deletion, as illustrated below:
(53) a. manhi-n kyoin-ka sul-lll mek-ko, manhi-nkyoin-ka
many-REL believers-S liquor-0 eat-and many-REL
nolim-lil hanta
believers-S gambling-0 do
'Many believers drink, and manybelievers gamble!'
b. manhl-n kyoin-ka sul-lll mek-ko hanta
(54) a. manhi-n hakca-ka Latine-lil malha-ko,manhi-n
many-REL scholars-S Latin-0 speak-andmany-41.EL
hakca-ka Latine-lll ss&nta
scholars-S Latin-0 write
'Many scholars speak Latir.,, and manyscholars write
Latin.'
b. manhIn hakca-ka Latine-lll Latine-lil ssInta
c. manhl-n hakca-kaLatine-lil malha-ko ssiAta
For these Korean data, one mightpostulate two different
'dialect' groupings: Group A consists of those speakers forwhom
only correferential NPfs aredeletable; Group Ell those for whom non- 25 correferential NP's can also be deleted and for whom, as a consequence, the (b) sentences of 53 and 54 may be ambiguous. For those in Group A, the (a) and (b) sentences of 53 and 54 are never synonymous. For both groups, 54b can be further reduced to 54c only if the NP's in question are correferential; the (a) sentences
7 can never be grammatical with correferential Nip's. In other words, for Group A, the (a) sentences have disjunctive, while the
(b) sentences have conjunctive, co-ordination. For those in
Group B, the co-ordination in the (a) sentences is disjunctive; the co-ordination in the (b) sentences, either conjunctive or disjunctive.For instance, 53a can only mean for either group that many believers either get drunk or gamble. Ex. 53b, however, can .10 mean for Group A that.many believers bothget drunk and gamble.
Both interpretations of 53b are possible for Group B, sothat this sentence can mean either that many believers either get drunk or
gamble, or that many believers both get drunk and gamble.
But the characteristics of 53, where Q is used, are also
found in 55 and 56, where true adjectives are used:
(55) a. pangthangha-n salam-ka sul-lll mek-ko, pangthangha-n
corrupt-REL people-S liquor-0 eat-and ,;orrupt-REL
salam-ka nollm-lil hanta
people-S gambling-0 do
'Corrupt people drink, and corrupt people gamble.'
b. pangthangha-n salam-ka sul-111 mek-konolim-lil hanta
'Corrupt people drink and gamble.' 26
(56) a. milyenha-n salam-ka khal-lIlkal-ko, milyenha-n salam-ka
foolish-REL people-S sword-0 sharpen-andfoolish-REL
khal-l11 hwitullnta
people-S sword-0 flourish
'Foolish men sharpen the sword, andfoolish men flourish
the sword.'
b. milyenha-n salam-ka khal-lilkal-ko khal-lil hwitulInta
'Foolish men sharpen the sword andflourish t1-11 sword.'
c. milyenha-n salam-kakhal-lil kal-ko hwitulinta
'Foolish men sharpen and flourishthe sword.9
(57) a. ki yeypp-n yeca-kasul-lll mek-ko, kl yeyppl-n yeca-ka
ithe pretty-REL girl-S liquor-0eat-and the prettpaEL
nolim-lil hanta
girl-S gambling-0 do
'The pretty girl drinks, andthe pretty girl gambles.'
b. )(I yeyppl-n yeca-kasul-lll mek-ko nolim-lli hanta
'The pretty girl drinks andgambles.'
(58) a. John-ka sul-lil mek-ko,John-ka nollm-lil hanta
'John drinks, and John gambles.'
b. john-ka sul-lilmek-ko hanta
'John drinks and gambles.'
Only when the subjectsof 56a are correferential can56b be further reduced to 56c for allspeakers. However, in 57 and 58, where definite and correferentialNP's are found, a disjunctive interpretation--that the prettygirl either drinks or gambles, or
- 27 that John either drinks or gambles--isimpossible for both sentences.
*But what causes so-called'dialectal, discrepancies like the above? The NP,s in question in 55 and56 are the kind that are called peneric here. The term generic seems tohave been used to denote two completely differentthings, as is illustrated in the following sentences:
(59) kolay-nn pholyutongmul-i-ciman,sange-nIn
whales-Tmammalian-are-but sharks-T piscine-are
'The whale is mammalian, butthe shark is piscine.,
(60) John-nln kh-ciman, Tom-nthcakta
John-T tall-but Tom-T short
'John is tall, but Tom isshort.,
(61) kolay-ka yeki-ey-to iss-ko,kolay-ka ceki-ey-to issta
whales-S here-too are-and whales-S there-too are
'There are whales here, andthere are whales theretoo.,
One use of generic denotes anentire class as opposed towhat -s not generics that are in the class. Contrary to popular belief, the
used this way are asdefinite as proper nouns(compare the subjects
of 59 and 60). We may say that kolay 'whale,is a name for the
whole class, as John is a namefor an individual. The other use of denoted by a the term refers to partof the membership of a class indefinitely used, as in common noun,in which case the generic is representation of an entireclass, 61. The first type, denoting a
will be hereafter termedrepresentative-generic; the latter,
denoting a representation ofpart of membership of aclass, 28 partitive-qcneric.8 Incidentally, this explains why Kuno (1970:1-4) has grouped definite NP,s and generic NPIsinto a single class, which can be foregrounded with the postposition wain Japanese. It
also explains why Carden (dh.1, n. 4) puts genericsinto the same
group as Ws. Kuno was referring to representative-generics,and
Carden, to both types of generic.
Returning to 55 and 56, if the subjectsof the (a) sentences
are representative-generic,naturally the sentences are synonymous
with the (b) sentences. Otherwise, if the subjects are not
correferential, conjunction reduction isimpossible. However, there
are some speakers of Koreanthat would allow conjunction reduction
even when partitive-generics arenot correferential, provided that
ambiguity is logically impossible--as in62-64. But for most
speakers, the (b) sentences of 62-64 areunacceptable as reduced
forms of the (a) sentences:
(62) a. manhi-n salam-ka cuk-ko, manhl-nsalam-ka pusang-11
many-REL people-S died-and many-RELpeople-S injury-0
ipessta
got
'Many people died, and manypeople got injured.,
b. *manhl-n salam-ka cuk-kopusang-111 ipessta
*,Many people died and gotinjured., 29
(63) a. yeypiA-n yeca-ka nolay-11pulless-ko, yeypp-n yeca-ka
pretty-REL girls-S songs-0 sang-and pretty-REL girls-S
piano-11 chiessta
piano-0 played
'Pretty girls sang songs, and prettygirls played the
piano.'
b. yeypr4-n yeca-ka nolay-11pulless-ko piano-111
chiessta
'Pretty girls sang songs and playedthe piano.'
(64) a. sey salam-ka cukess-ko, seysalam-ka pusang-111 ipessta
three men-S died-and three men-Sinjury-0 got
'Three men died, and three mengot injured.'
b. *sey salam-ka cukess-kopusang-111 ipessta
*1Three men died and gotinjured.'
To sum up, conjunctionreduction is possible only whenthe subject NP's are correferential.When the correferentialityof two
NP's is not explicitlyindicated either positively ornegatively--
i.e. when a Q7 a numeral, or adescriptive adlective is the
modifier of a [-specifid] NP--a reduced sentence has theconjunctive
reading of its predicates(e.g. 53b and 54b). But an unreduced
sentence has the disjunctivereading of the predicates(e.g. 53a inclusive and 54a). The disjunction in the latter casecan be either
or exclusive. If exclusive, the reducedand unreduced sentences
cannot be synonymous. If inclusive, they may be synonymous.
2.15. Questions. Carden (48) claimed that onlywhen a
pre-determiner Q is involved, as in65b, does an interrogative n 30 sentence question a modifier. Otherwise, it questions the highest predicate of the sentence.. Butwhat about ex. 661 where a non-Q adjective modifies theindefinite NP?The modifier of girls is a true adjective, viz, pretty,and yet the sentence cannotquestion the highest predicate, apply,without also questioning the descriptive adjective:
(65) a. Did the cholera patientssurvfve? (Questions survive.)
b. Didmgny patients survive? (Questions many.)
c. Did the manypatientssurvfve? (Questions survive.)
(66) Did prtty girls apply?
In Korean, markedunexpected) and unmarked (normal)accent
indicates what part of asentence is questioned. Thus:
(67) a. kg khollela hwanca-t11-kasalanamass-o? (Matked accent)
the cholera patient-PL-Ssurvived-QUES
'Was it the cholera patientsthat survived?'
b. ki khollela hwanca-tl-kasglanamass-o? (Unmarked accent)
!Did the cholera patientssurvive?!
(68) a. kg yeyppl-nyeca-t411-ka Ingmohayess-o? (Marked accent)
the pretty-REL girl-PL-Sapplied-QUES
!Was it the pretty girlsthat applied?' e b. kl yeyppl-n yeca-tll-kawigmohayess-o? (Unmarked accent)
!Did the pretty girlsapply?'
(Marked accent) (69) a.yeeypp&-n yeca-til-kaingmohayess-o?
'Were the girls whoapplied pretty/1
b. yeyppl-n yeca-til-ka&ngmohayess-o? (Unmarked accent)
'Did pretty girls apply?' (Questions apply.) 31
(70) a. manhi-n yeca-tl-kangmohayess-o? (Marked accent)
many-RFTJ girl-PL-S applied-QUES
'Was it many girls thatapplied?'
b. cmanhi-n yeca-.U1-kaLgmohayess-o? (Unmarked accent)
'Did many girls apply?' (Questions apply.)
From the descriptions sofar given, the reader willprobably indexes on have gathered thatcorreferentiality--i.e. identical
specifiC] NP's--is the determining factor in EQUI-NPDELETION and 11-
CONJUNCTION REDUCTION, wheretwo or more NP's areinvolved, whereas a differencein definiteness is asufficient condition for Since different interpretationswhen only one NP isinvolved. sentences 67-70 have only onesubject, the conditioningfactor in
show, in the (a) them is definiteness. As the translations
sentences of 67 and68--where the subject NP's aredefinite--both (marked and (b) and (c) readings arepossible with the different subject NP's are indefinite, unmarked) accents.In 69 and 70, whose talk about only the (b) readings arepossible, since one cannot either the physical something whose existenceis not presupposed in
or an imaginaryuniverse.
The derivation ofindefinite NP's modifiedby adjectives. to prove .2. So far, we have usedCarden's own arguments Q's are not that the anomalouscharacteristics of pre-determiner
peculiar to Q's but aretrue of all indefiniteNP's modified by higher-predicate adjectives. It has just been assumedthat Carden's to accept analysis was correct forthese. But it would be well not
41 32 this too hastily, in the absence of a search for counterevidence.
There seem to be two different classes of predicate adjectives: (A) those which can have both definite and indefinite subjects; and (B) those which can have only definite subjects.
(71) a. *Man is many.
b. *Men are many. 9 c. Those men were many.
d. The men who came were many.
(72) a. Man is tough. 10 b. (Certain) men are tough.
c. Those men were tough.
d. The men who came were tough.
Many, five, and the like are of the second group;and tou2h, patty,
and so on are of the first group.In other words, definiteness in
a noun has to be establishedby some previous sentence for a (B)
adjective to be used with that noun, asillustrated by 73 below:
(73) a. (Certain) men died; and the men were many.
b. *(Certain) men were many; and theydied.
Suppose 73a is stated by X and negatedby Y, as in 74.
(74) X: (Certain) men died; and the men whodied were many.
Y: You are wrong.
Since we are dealing with two conjuncts, we canthink of three
logical possibilities for the scopeof negation: negation of the
first conjunct only, negation of the secondconjunct only, and
negation of both conjuncts. But as shown below, one possibility
42 33
(viz, a) is in fact logicallyruled out:
(75) a. *It is not true that(certain) men died; and the men
were many.
b. (Certain) men died; andit is not true that the men were
many.
c. It is not truethat (certain) men died; andit is not
true that the men were many.
Ex. 75a is inconceivablebecause the ground for theproposed
definiteness of the second occurrenceof men is absent.The
negation of the firstconjunct entails theimpossibility of the possibilit5erl, we are second conjunct. Out of three mathematical neqa'zion of the left with only two: negation of both conjuncts, or made for (A) adjectives; second conjunct. Similar arguments could be
note, therefore, thefollowing: behind. (76) X: (Certain) men got hurt; andthose men were left
Y:You are wrong. true; (77) a. *The statement that(certain) men got hurt is not
and those men wereleft behind.
b. (Certain) men gothurt; and !t is not truethat those
men were leftbehind. and it is not c. It isnot true that(certain) men got hurt;
true that those men wereleft behind.
existence of the men isnegated Ex. 77a isungrammatical since the existential assertion in the first sentence. In other words, if any following conjuncts do not of an NP is negatedin a conjunct, the
4 3 34 contain verbs that have the NP astheir subject until the existence of the NP has been establishedagain. This seems to present itself as a strong candidate for auniversal aspect of language. Exactly the same patterns as in Engli-. are foundin Korean:
(78) a. (etten) salam-til-ka manhassta
certain man-PL-S were-many
*1(Certain) men were many.'
b. yeki-ey o-n salam-til-ka manhassta
here-to came-REL man-PL-S were-many
'The people who came here were many.'
c. ki salam-til-kamotu-iessta
the man-PL-S were-all 111 ?,The people were all.
(79) a. (etten) salam-tll-nnkanghayessta
certain man-PL-T were-strong
1(Certain) men were strong.'
b. yeki-ey o-n salam-tll-kakanghayessta
here-to came-REL man-PL-S were-strong
'The men who came here were strong.'
c. kl salam-tll-kakanghayessta
the man-PL-S were-strong
'The men were strong.'
In Korean, in contrastwith English, the use of the
definitizer, kl 'the', is limitedto situations wheredefiniteness
is conditioned byovert'expressions (i.e. not entailed by an 44 35 unexpressed presupposition). Kartunnen.(1971:29-30) lists the following as the non-linguisticenvironments conditioning the use of the definite article in English(the numbering is mine):
(80) Anyth'.1g in the immediate environmentof the speaker and the .learer towards which theirattention is directed becomes a discourse referentwhether it has been explicitly mentioned or not.
(81) In every discourse, there is abasic set of referents which are known to exist althoughtheir existence has neither been asserted nor observedduring the discourse itself. This set is determined by the commonunderstanding the participants believe theyshare with regard to their environment.
(82) A discourse referent is establishedwithout any explicit introduction, provided that its existence canbe inferred with some degree of certaintyfrom the existence of another referent by an implication withwhich the listener supposedly is familiar. The status of the derivedreferent is the same as that of itspremise.
The above general characterizations areillustrated below by the
sentences which are identicallynumbered:
(801) a. 1. The roof is leaking.
2. Look at the picture onthe wall.
3. Fetch him out of the bed.
D. 1. cipung-ka saynta
roof-S leaks
'The roof leaks.'
pyek-wi-y k1im-11 po-la
wall-on-of picture-0 look-IMP
3. k1-11 chimtay-eysekklcipenay-la
him-0 bed-from fetch-out-IMP 36
(81') a. 1. Man can (now) go tothe moon.
2. He went to see.the doctor.
3. Mary went to the beach.
b. 1. inkan-ka tal-ey kanta
man-S moon-to goes
2. Id-nki 1122-eykey kassta
he-T doctor-towent
3. Mary-nin laa_sccan-eykassta
Mary-T beach-to went
(821) a. 1. I was driving on thefreeway the other day when
suddenly the engine began tomake a funny noise. I
stopped the car, and when I openedthe hood, I saw
that the radiator was boiling.
b. 1. yocennal kosokto10-2A1tallinn-tey kapcaki eyncin-ka
the-other-day highway-0 was-running-andsuddenly
isangha-n soli-lil nay-ki-sicakhayessta.cha-11
engine-S strange-REL noise-0make-NOM-began car-0
mec-ko hus-11 ye-ni latieytha-kakkIlh-ko-issessta
stopped-and hood-0 opened-and radiator-Swas-boiling
Note that k& is not addedbefore any of the underlined NP'swhich were definitizedin English.None of the above threegeneralizations by Kartunnen is validfor Korean.
In the previous section,definiteness was defined as what is
assumed to be regist,T-ed inthe knowledge of both thespeaker and
the hearer. Thus, in an example like83a below, the existence of 3 7 the book is already known to thespeaker--in our terms, it is specific--but not yet known to the addressee. After 83a is said to the person addressed, the existenceof the book is also registered in his ruind, and is therefore definite.Correferentiality has been defined as what is identically-indexed and[+ specifiC].
Indefinite NP's can be correferential, as is shown bythe reflexivization and pronominalization found in 83band 83c, respectively:
(83) a. I bought a book yesterday; and I lent thebook to John.
b. A man killed himself last week.
c. A man killed hiswife in the town today.
The above discussion has made clear, Ihope, the differences
between Korean kand English the which cause so manymistakes in
English when it is spoken by nativespeakers of Korean, and which
explain why English speakers putsuperfluous ki.'s in their Korean.
To reiterate, Koreans do not usekexcept when a registering in
the knowledge of the hearer isovertly caused by the speaker,
either by previous utterances or by deicticexpressions (e.g. 84b
and 84c). Thus, the underlinedwords below are definite: the first
is definitized by a previouslinguistic utterance; and the rest, by
deictic expressions, most probablywith pertinent gestures:
(84) a. nay-ka ecey chayk-ll sassnln-tey, k&chayk-2A1 John-ka
I-S yeSterday book-0 bought-and the book-0 John-S
pillyekassta
borrowed-from-me
bought a book yesterday, and Johnborrowed it from me.' 38
(84) b. i chayk-lil ilke-la
this book-0 read-IMP
'Read this book.'
c. ce chayk-lIliii kaceo-la
that book-0 here bring-IMP
'Bring that book here.'
Both i 'this' and ce 'that' areused when the indicated items are
present to both speaker andaddressee. The particle i is used with
reference to closer items, while ceis used for more distant ones.
The definite-marker k& 'the' seemsto be possible only when an NP
has been overtly registeredin the knowledge of the addresseeby the 12 speaker through fome previouslinguistic expression. In the next
chapter, I dis-__ c!xt..lisivelyhow kl can be considered to bethe
pro-form of a 1:c.::t:i.;!:ive'relativeclause. I assume that
Kartunnen's generalizations asto definitization are universal.
For although definite NP's arenot ovettly marked in Koreanoutput
sentences, e definitenessof NP's is revealed inrelativization
and topicalization.Returning to 78 and 791 theungrammaticality
of 78a results from 'thefact that manhta lbe manylis a (B)
adjective and thel.efore cannot have anindefinite NP as its subject.
Even though overt markingof definiteness is a language-
particular matter, as is clearlyJemonstrated by the difference
between what is found in Kor.: and ,:hat is found in English,the
.. : verb feature definite:t j seems to be universal.Thus, [C-1 SZNIMMIMelll 85b is ungrammatical forthe same. Lnat 73b is ungrammatical: 39
of the first occurrenceof any (B) adjectivescannot be predicates is supplied byextra- NP unless theground for definitization not use (B) linguistic means, asin exs. 80-82. One simply does establish the 2xistence adjectives as predicatesof sentences which
of the particularitem in question. kilentey k& salam-t4l-ka (85) a. (etten)salam-til-ka cukessta; the man-PL-S certain man-PL-S died and
manhassta
were-many
1(Certain) men died;and the men weremany.' klentey ksalam-tC.-ka b. (etten)salam-til-ka manhassta; the man-PL-S certain man-PL-S were-manyand
cukessta
died
1(Certain) men were many;and the men died.'
negating 85a: either There are only twopossibilities of
statement or that ofthe secondconjunct, the negationof the whole first. Thus, 86a is synonymous liut never thenegation of only the
with 86c, but neverwith 86b below: salam-til-ka (86) a. (etten) cukess-ko k
certain mm-PL-S died-andthe man-PL-S ani-ita manhassta-ko-ha-ciman, klkes-nln sasil-ka fact-S not-is were-many-QT-say-butthe thing-T and the men were 'It was said that(certain) men died
many; butit is not true.'
40 40
(86) b. (etten) cukessta-ko-ha-nin-kes-nIn sasil-ka
certain man-PL-S died-QT-say-REL-thing-T fact-S
ani-ita; kilentey ksalam-t11-ka manhassta
not-is and the man-PL-S were-many
'It is not true that (certain) men died; but the men were
many.'
c. (etten) salam-til-ka cukessta;Iclentey kl salam-til-ka
certain man-PL-S died and the man-PL-S
manhassta-ko-ha-nln-kes-nin sasil-ka ani-ita
were-many-QT-say-REL-thing-T fact-S not-is
1(Certain) men died; but it is not true that the men
were many.'
Ex. 86b is ungrammatical for the reason that is givenabove. I will naw turn to justifying what I havemerely described in section 2.1.
2.21. Other ne ations and questions. Carden recognized the intricate relations between definiteness of subject NP's and Q's, but did not notice that what I am calling (A) adjectivesalso behave in the same way.He recognized, in effect, that when Q's occur before the determinerthey are not only indefinite but non- specific. Q's, which are the predicates of a higherunderlying
sentence according to the Lakr,ff-Carden hypothesis,should be the
predicates of definite subject-NP's. This wiLL be illustrated in
87b, the logical transckiption of 87a below:
(87)a.(Certain) men left; I do not think that the men were
many.
b.((Certain) men left) . (I think t(the) men were many] ) 50 41
According to De Morgan'sTheorems, -(p q) a .p v .q.
So far as the truthvalue of a stetement isconcerned, we are left with three possibilities: -A, or (,.-p But I have already discussed how implies (,--p ,g) when the occurrence of is an NP establishesits existence in thefirst conjunct and the NP repeated in the followingconjunct. Then:fore, there are only two possibilities of negating asentence like 74: either negating both conjuncts,as in 75c,or negatingonly the second conjunctoasin 75b.
Sentence 75c, however, isimplied by 75a,whic!h negates onlythe 13 counterexamples to what first conjunct. But interesting apparent
are given inKartunnen (10) are quoted as88 below:
(88) a. Peter said that Johndid not catch a fish andeat the
fish for supper.
b. George did not steal anapple and eat it.
c. John didnot buy a bracelet, nordid he give the
bracelet to Mary.
negative markers These sentences lookanomalous because they have negation of the first on both conjuncts,despite the fact that the
conjunct should implythat of both conjuncts. In other words,
negation is redundantlymarked in 88. This is, however, not a is used only when real counterexample,since this redundant marking of which have been one is negating aconjunction both conjuncts
claimed by others. This is not even exceptional,for each conjunct Only any conjunctionwould have to carry NEG to benegated:
the first coniunct of89a is negated in 89b, justbecause NEG occurs 42 only in the first conjunct; both conjuncts are negated in 89c simply because NEG occurs in both conjuncts in that example.
(89) a. John likes reading; and he dislikes swimming.
b. John does not like reading; and he dislikes swimning.
c. John does not like reading; but he does notdislike
swimming.
If the negation had not been redundantly marked, 88 would have haen ungrammatical for t'ne same reason that makes 75a ungrammatical.
Another difficulty yith the proposals being put forward here arises when the correfrential NP's in the conjuncts are all definite, as in 90a. This has been correctly described by
Carden (56) as derived from 90b. In the following sections, I hypothesize that all the embeddings under discussion are derived from underlying conjunctions. In examples like 90, q) is
equivalent to ,..10 v q v .40; p does not imply (..,p .40.
Why then is it thctt the NEG attached to a NT-verb, think, negates
only left in the following examples?
(90) a. I don't think that the five men left.
b. I don't think that the men, who were five (in number),
left.
Two solutions, which are not mutually exclusive, suggest
themselves. One explanation would be that most, if not all, non-
restrictive relative clauses are outside the tree containing the
matrix sentence. In other words, they are mostly interpolations
by the speaker, almost like a performative declaration.According
to this proposal, 90b would have 91a as its deeper structure.The
623..tw , 43 logical transcription wouldbe 91b:
(91) a. I don't think thatthe men left; and the men werefive
(in number).
b. (I think[,-Lthemenlef41).((the)men five (innumber))
To make this pointclearer, I would like tostate what probably has been always assumed: all restrictive relativeclauses definitize 21E221y definite NP's may be head nouns. To put it differently, no only modified by restrictiverelative clauses. So 92 could be used when two or more John's arepresupposed to exist in theshared knowledge of speaker andhearers:
;92) I am not talking aboutthe John who is Mary'sbrother, but
the John who is Tom'sbrother.
According to this analysis,all the restrictiverelative clauses existence of the should have come from thefirst conjunct, where the
head noun is announced.Such a conjunct will benamed anintrolustoa which introduces sentence (abbr.intro-S). An intro -S is a conjunct be the existence of afact or a thing. Nctice that 93a could
derived from 93b, but notfrom 93c.
(93) a. The men who have comefrom Hawaii are successful.
b. (Certain) men have comefrom Hawaii; and the men are
successful.
c. The menhave come from Hawaii;and the men are
successful. conjunct or In negating 93a, onecould negate her the second conjunct in the both conjuncts of 93b,but never only the first 53 44
'sentence. It is clear from this that when aconjunction the iirst
conjunct of which is an intro-S iswithin the scope of a NEG, that
NEG cannot negate the first conjunctwithout negating the second
conjunct also. If the embedded complement includes a non-
restrictive relative clause, the latter isusually outside the
scope of the NEG which standsabove the matrix verb.
The second explanation would beclosely related to what has
been stated in the previous section: In a sentence with a definite
subject, the accented part of thesentence--whether subject or
predicate--is negated by the NEG above thematrix verb. Thus, to
make la synonymous with lb, leftin la should be accented; to make
it synonymous with lc, attiactivethere should be the most heavily
accented part. The sentences of 1 are repeatedhere for convenience:
(1) a. John doesn't think theattractive girls left.
b. John thinks the attractivegirls stayed.
c. John thinksthe unattractive girls left.
This reveals a very interestingaspect of negation:NEG negates the
most heavily accented part of thesentence within its scope. This
characteristic of language eliminatestwo possibilities in the
negation of embedded conjunction. Above, we noted that "-(p q)
is equivalent to p v ,wq v 0....13 In the negation of an
embedded sentence including anon-restrictive relative clause
(abbr. NR), the negationof the conjunct which representsthat NR-- conjunct is buried either p or not possible, because that
in the subject NP and shelteredfrom the reach of negation, soto
speak. 45
modifiers of This gives us a reasonfor believing that derived from an underlying embedded indefinitesubje=s are to be since only in thatevent second conjunct, asshown in 94 and 95, adjectives not contradicted. is the previousexplanation of (B) of adjectives paralleleach (Seebelow for ways inwhich both types
other.)
(94) a. Many girlsapplied. many. b. (Certain) girlsapplied; and the girls were
(95) a. Pretty girlsapplied. girls were pretty. b. (Certain)girls applied; and the
222.1:121 as their verb. Here, a rule The intro-Sysin 94 and 95 have from 94b and 95b,respectively, is needed toderive 94a and 95a pieces of information,and a since the (a)sentences carry two of information inthe deep single sentencecarries only one piece formulated immediatelybelow seems to structure. The rule to be conjunct refers to operate only whenthe NPi in the first denoted by that NP,i.e. it is unspecified membersof the universe constrainf on the a partitivegeneric. There is a further the second conjunct application of rule96: the predicate of Now, the rule is must be a simpleattributive adjective.
formulated as 96; the marked(m) stress of a NP is afalling contour. FORM)14 (96) NR (Adj-Pr)FORMATION (abbr. NRrAdj-Prj
(a, NP. 7 x) 17 Si 2 definitel<-1- presupposed> specific) 1 2 3 4
>1, 4 + 2, 3, V
So far, I have explained why the NEG immediatelydominating a sentence whose subject isindefinite and is modified by 6n adjective negates the adjective (e.g. many in 3a andattractive in
15a) but not the predicate of the embedded complement(e.g. left in
3a and 15a). Ex=1.ct1y the same arguments apply to implicit negation 15 and questions. Consider the following:
(97) a. John denies that many candidates won.
b. John denies that rich candidates won.
c. John denies that the richcandidates won.
(98) a. Did many candidates win?
b. Did rich candidates win?
a c. Did the rich candidates win?
Simplified logical interpretationsfollow:
(97f)a. ((Certain)candidateswon)-(John deny ((the) candidates
many))
b ((Certain)candidates won)-(John deny ((the)
candidatesrich))
c. ((Certain)candidates rich)-(John deny ((the)
candidateswon))
(981) a. ((Cee-ain)candidates won)-(I ask you ((the )
candidatesmany))
b. ((Certain)candidates won)-(I ask you ((the)
candidatesrich))
c. ((Certain)candidates rich)-(I ask you ((the)
candidateswon)) 47
then Hlesucceeding If we deny orquesbion an intro-S, the definitenessof the NP1s sentences becomeungrammatical, since have to beunmotivated. At this in the secondconjunct would now RELATIVE CLAUSEFCRMATION (abbr. point, I willintroduce RESTRICTIVE in greater detailin RR-FORM), a rule whichwill be discussed intro-S into aconjoined following Chapter Three. RR-FORM copies an This rule movesin thd sentence whichshares an NP withit. The (c) sentencesof 971 opposite directionfrom NR (Adj-Pr)FORM. In 981c, therelativized and 981 will berelativized by RR-FORM. into a questionafter the sentence will befurther transformed Rule 96 willderive the (a) deletion of theperformative clause, from the respectivesentences of 971 and (b) examplesof 97 and 98 derive the (a)and (b) of 97"and 98": and 981; andRR-FORM will that won were many. (97") a. John deniesthat the candidates that won wererich. b. John deniesthat the candidates many? (98") a. Were thecandidates that won
b. Were thecandidates that wonrich? the Koreanversion et I will nowattempt to fomulate 12, 13, and 14,whie; are NR (Adj-Pr)FORM.Let us reconsidr
repeatd belowfor convenience: sayngkak-ha-ci (12) kyeypiA-n yeca-tll-kanathanassta-ko-nln think-NOM the pretty-RELgirl-PL-Sappeared-QT-DLM
ani-hanta
not-do appeared.' 1I do notthink that thepretty girls 48
ani-hanta (13) manh-n salam-t11-kakassta-ko-nln sayngkakha-ci not-do many-REL man-PL-S went-QT-DLM think-NOM
do not think that many menwent.' ani- (14) yeyppi-n yeca-t_U-kanathanassta-ko-nth sayngkakha-ci not- pretty-REL girl-PL-Sappeared-QT-DLM think-Nr)M a hanta
do
do not think thatpcetty girls appeared.' along the lines ofthe Simplified logicalinterpretations would be following: yeca--U1 (121) ((etten) yeca-t11yeyppessta).(C: k&
certain girl-PL were-pretty the girl-PL
nathanasstaH sayngkakhanta)
i appeared think i _ manhassta)..)) 1 (131) ((etten) salarn-t1kassta)-((( ksa1ara-t11 1 certainman-FL went the man-PL were-many 1
sayngkakhanta
think
(141) 1:(etten) yeca-tllnathanassta)-(((k1 yeca-t11
certain girl-FL appeared the girl-PL
yeyppessta)...) sayngkakhanta)
were-pretty think surface The rule utdchwould turn these intotheir respective as 96. Sdnce all structures, viz.12-14, canbaagain formulated surface relative Koreanattributive modifiersof NPts are actually
17c 49 clauses, item 4 of 96 hasto be relativized in Koreanwhen transported ahead of item 2.However, this Korean versionof
NR (Adj-Pr) FORM shouldnot be confused withRR-FORM; the two rules move in oppositedirections.Exs. 13' and 14', fromwhich
NR (Adj-Pr) FORM derivesthe complements of 13 and14, respectively, are tran-Sformedinto 13" and 14".byRR-FORM:16
(13") ka-n salam-t11-ka manhassta
went-REL man-PL-S were-many
'The men who went were many.'
(14") nathana-n yeca-t11-kayeyppessta
appeared-REL girl-PL-Swere-pretty
'The girls who appeared werepretty.'
As a result of 2.22. EOUI-NP DELETION andREFLEXIVIZATION. have been able to describe our definitionof correferentiality, we
EQUI-NP DELETION andREFLEXIVIZATION adequately(cf. 2.13 and 2.14).
Consider now thefollowing sentences and thelogical interpretations what has been said which are easilyposited along the lines of
already: be admitted. (99) a. Many studentsexpected many students to (((certaix) students. be
((Certain)students.expected -
admitted)-((the) students. bemany)))-((the students .
were many)
b. The students whoexpected many students tobe admitted
were many.
c. Many studentsexpected the students whowould be
admitted to be many. 59 50
(99) d. The students who expected the students who would be
admitted to be many were many.
e. Many students expected tobe admitted.
((Certain) students. expected ((the) students. be 3- 3- admitted)).((the) students, were many) 3-
( 10 0 ) a. Democratic candidates expected Democraticcandidates
to win.
((Certain) candidates. expected ( ( (certain) candidates. 3- 3 win)( (the) candidates be Democratic) ) )( (the)
wercandidates. e Democratic)
b. The candidates who expected Democraticcandidates to
win were Democratic.
c. Democratic candidatesexpected the canadates who would
win to be Democratic.
d. The candidates who expected the candidateswho would
win to be Democratic were Democratic.
e. Democraticcandidates expected to win.
((Certain) candidates. expected ((the) candidates. a. 3- win)) ((the) candidates. were Democratic) 3- ( 10 1) a. John expected John to beadmitted.
(John. expected(John be admitted)) 3- b. John expected to be admitted.
(John. expected (John be admitted)) 3- In the (a) settences of99 and 100the existence of the subject
DIP s of the complement sentences wasnewly introduced in the first 51 conjuncts, which would havebeen impossible had the NP'sbeen correferential with the matrixsubject NP's of the first major conjuncts. In the (e) sentences,however, all subject NP's except the first NP modified by certain aredefinitized by the matrix part of the first conjunct, anddefiniteness is not extra-linguistically conditioned; as a result, they arecorreferential. These exactly correspond to 101a and 101e,respectively.
Furthermore, since both NR (Adj-Pr) FORMand RR-FORM are clearly optional rules, we havenine possibilities of combination when two double-conjunctconjunctions are considered.Only four of them will be dealt with here,since each of the other fivedepends on thenon-application of either or both ofthese two rules.The
four combinaticns to bedealt with here are: (i) NR (Adj-Pr) FORM
NR CAdj-Pr) FORM;(ii) NR (Adj-Pr) FCRM + RRFORM;(iii) RR-FORM 4,
NR (Adj-Pr) FORM; and(Iv) RR-FORM + RRFORM.The (a), (b), (c),
and (d) sentences of 99 and100 are the (i), (ii), (ill),and (iv)
cases, respectively.Since all of them are from the sameunderlying
structure, they are synonymous. The (b), (c), and (d) sentences
all show that the identityof the NP's in the (a)sentences is
accidental; in other words, theNP's are not necessarily
conterminous sets, but may be soin a given situation. Accidentally
identical NPfs are notcorreferential and therefore arenot subject 17 to EQUI-NP DELETION. The -same arguments alsoapply .to
REFLEMEMATION Let vs consider thefollowing: 52
(102) a. Many men love many men.
((Certain)men.love(certain)men.)-((thOmeniare
mny) ((the) rnen. are many)
b. The men who many men love are many.
c. The men who love many men are many.
d. *The men who the men who love are many are many.
e. Many men love themselves.
((Certain)ren.loverren.)*((the) men1 are many)
(103) a. Haughty men hate haughty men.
((Certain)rell.hate(certain)men.)-((thOmeniare
haughty) -( (the) men are haughty)
b. The men who haughty men hate arehaughty.
c. The men who hatehaughty men are haughty.
d. *The men who the men who hate arehaughty are haughty.
e. Haughty men hatethemselves.
((Certain) men. hatemen. ) -((the) men1 arehaughty)
(104) a. John loves John. (x loves y)
b. John loves himself. (x loves x)
In 102a and 103a, the existenceof both underlined NP's in the first conjunct is newly introduced: they are not correferential with each other. In 102e and 103e, however,all the NP',s except the first one are definite becauseof the prior occurrence of the NPin the first conjunct; this shows that they arecorreferential. We see a direct parallel between these and therespective sentences in 104. 53
Even though we have three conjuncts conjoined in 102a and
103a, we have only two possible conjunctions of two conjunctseach:
conjunction of the first and the second conjuncts, and conjunction
of the first and the third conjuncts. Conjunction of the second
and the third conjuncts is impossible, since the NP's ofthe two
conjuncts are not correferential with each other.Again, we have
four possible combinations for the application of NR(Adj-Pr) FORM
and/or RR-FORM: the (a), (b), (c), and (d) sentences of 102 and
103 are the (i), (ii), (iii), and (iv) cases describedabove,
respectively.The ungrammaticality of 103d is accounted for by
Ross's (1967:66-88) Complex-NP Constraint. The ungrammaticality of 06 102d is also caused by the fact that a Q cannot be thepredicate of
an intro-S and consequently cannot bethe predicate of a restrictive relative clause.
The above examples seem to show the correctnessof the
assumption that NR (Adj-Pr) FORM and RR-FORM are notcrucially
ordered with respect to each other.This also appears to be the
case in Korean. Consider the following:
(I05) a. manhl-n haksayng-t&l-ka manh-nhaksayng-tU-ka
many-REL student-PL-S many-REL student-PL-S
suyong-tcylila-ko kitayhayssta
be-admitted-QT -expected
'Many students expected, many students ,to beadmittea.1
((etten) haksayng-t&li kitayhayssta (CCetten) haksayng-
til. suyong-toyta),(Ck&) haksayng anhta))) C(kl) 3 , haksayng ,manhta) 63- 54
(105) b.manhl-n haksayng-ti1-ka suyong-toy-ki-111 kitayha-n
many-REL student-PL-S be-admitted-NOM-0 expected-REL
haksayng-t11-1An manhassta - student-PL-T were-many
'The students who expected many students to be admitted
were many.'
c. manhl-nhaksayng-tia-ka suyong-toy-1 haksayng-t11-ka
manh&lilako kitayhayessta
'Many students expected the students who would be
admitted to be many.'
d. suyong-toy-1 haksayng-til-ka manhIlila-ko kitayha-n
haksayng-t11-nln manhassta
1The students who, expected the students who would be
admitted to be many were many.'
e. manhl-n haksayng-tIlka (caki-ka) suyong-toylila-ko
kitayhayessta
'Many students expected to be admitted.'
((etten) haksayng-tIli kitayhayessta ((kl) haksayng-tIli
salyong-toyta)) ((kAt) haksayng-t111 manhassta)
(100 a. akha-n saIam-tIlka akha-n ton-1-11
wicked-REL man-PL-S udcked-REL man-PL-S money-0
kitayhayessta,
make-14011-S, expected
'Wicked men expected w±cked men-to mak-emonel.f
((etter) salam-t11. kitayhayssta Metten) salam-t11_ ton 3 pelta) ( (WO akhata) ) )( CkI) salam-tfl, akhata) 55
(106) b. akha-n salam-til-ka ton-lil pel-ki-lil kitayha-n
wicked-REL men-PL-S money-0 make-NOM-0 expected-REL
aknayessta
man-PL-T were-wicked
'The men who expected wicked men to make money were
wicked.'
c. akha-n salam-til-ka ton-lil pel-nin
akha-ki-111 kitayhayessta
'Wicked men expected the men who would make money to be
wicked.'
a. ton-lil pel-nln salam-til-ka akha-ki-Alkitayha-n
salam-t&l-nin akhayessta
'The men who expected the men who would make moneyto
be wicked were wicked.'
e. akha-n salam-t&1-ka (caki-ka) ton-111 pel-k1-111
kitayhayessta
'Wicked men expected to rake money.' ((etten)salarayhayessta ((k1) salam-til ton
pelta))-((k1) salam-t&l. akhayessta)
The (a) -(d) sentences are all synonymousin 105 and 106, and are different from the respective (e) sentences. These exactly parallel the correspondina English sentencesof 102 and 103. It should be clear at this point wtyREFLEXIV1ZATION(except the inclusive type, where appalcable) is possibla =the(e) sentenc 5 6
2.23. CONJUNCTION REDUCTION.Let us consider 45 and 46 together with 107: (45)a.Goodrules.areexplicit;andgoodrules.are easy to
read.
b. Good rules are (both) explicit and easyto read.
(46) a. Few (of the) rules are explicit; andfew (of the) rules
are easy to read.
b. Few (of the) rules are (both)explicit and easy to read.
(107) a. Pious men ere good sons; and pious men aregood fathers.
b. Pious men are both good sons andgood fathers.
In 45, the NP's are what I have earliertermed representative-generic; they are therefore definite and,consequently, correferential. In 46 and 107, however, we are talking aboutpart of the membership of the group(partitiVe-qeneric), as shown in the logical formalism below: (461)a...(((certaisules.explicit)((the) rules. few)).
(((certain)rules.easytoreacp((the)rules.few))
b. (((Certain) rulesexpaicit)-((the) ruleseasy to
read))-((the)rules.are few)
(1071) a. (((Certain) meni goodsons).((the) meni pious)).
(((certain) menj good fathers)-((the),menipious))
b. (((Certain) meni goodsons)-((the) 'went good fathers)).
((thOmen-lolous)
In other words, the existenceof the NP's is introduced in both major conjuncts of the (a) sentences,while all of the subject NP's are definitized bythe first occurrence of the NP inthefirst 57 conjunct of the (b) statements.Therefore, the predicates are conjunctively co-ordinated in the(b) statements but disjunctively co-ordinated in the (a) statements. If shared NP's in conjunction
are representativegenerics, they are necessarilycorreferential.
Let us return to data fromKorean:
(108) a. manhl-n salam-t11-kacuk-ko manhi-n salam-til-ka
manpaEL man-PL-S died-and many-REL man-PL-S
pusang-111 ipessta
injuryO got
'Many men died, and many mengot injured.'
Metter) salam-til.cukessta)-((k4) sa1am-t11.
manhassta)) -(((etten)salam-t11.pusangipessta)- 3 ((k) salam-til3 .manhassta))
b. smanhl-n salam-t11-kacuk-ko pusang-lil ipessta
?"Many men died, and gotInjured.'
(((etten)salam-t11.cWmssta)-((kl)salam-t11pusang
ipessta))-((kit) manhassta)
(109) a. yeypp4-n,yeca-tIl-kanolay-lil pulless-ko,yeyppl-n
pretty-REL girl-PL-S song-0 sang-and pretty-aEL
yeca-til-ka khyessta
girl-PL-S violin-0 played
'Pretty girls sang songsand pretty girls playedthe
violin.'
Metter) yeca-t&l.nolaypuIlessta)-((ki) yeca-tili
yeyppessta))-((f.etter) yeca-t1.1. khyessta)-- J ((kA) yeca-t11.yeyppessta)) 58
(109) b. yeyppl-n yeca-t11-ka nolay-111 pull-kovio1in-111
khyes sta.
*( ( (etten) yeca-tAli nolay pullessta) ((k1)yeca-tIli
violin khyessta) )((ki) yeca-till yeyppessta)
Although the deep structures in 108b and109b are logically
impossible, their very logical impossibi lity seemsto allow at least
some speakers to usethe surface structures In 108b and 109b as
reduced forms of the (a) sentences. In other words, when the (b)
sentences of 108 and 109 are used, they areunambiguously understood
as the reduced formsof the respective (a) sentences. Some of, these
speakers seem to have pushed thisgeneralization of CONJUNCTION
REDUCTION a little further, so that evenwhen ambiguity is
possibleas in 110 belowconjunctions canbe reduced.Thus, in the
speech of those people 110b can meaneither what is given in the
logical interpretation of 110a, orwhat is given in that of 110b:
(110) a. manhl-n ,kvoin-t11-kasul-lil mek-ko, manhl-n kyoin-
manyREL believer-PL-S liquor-0drink-and manyAtEL
tIl-ka toplak-111 haaba
b,ellever-PL-S gambling-0 do
/Many believers drink, andmany believers gamble. ' 59
However, when a subject NPrefers to a group witha smaller membership (cf. 111), it is moreeasily specifiable, and consequently the generalization ofthe CONJUNCTION-REDUCTION rule seems to be prohibited in thespeech of all speakers.Thus, no speaker regards llib as synonymouswith llla:
(111) a. myet kyoin-til-ka sul-lilmekess-ko, myet kyoln-t11-ka
a-couple believer-PL-S liquor-0drank-and a-couple
topak-11l hayessta
bellever-PL-S gaMb1ing-0 did
fA couple of believers drank,and a couple of believers
gambled.'
Metten) kyoln-till sul mekessta)((kl) kyoin-tIli
myet-iessta))-Metten) kyoln-tll, topakhayessta)- J ((kA) kyoln-tllmyet-iessta))
b. myet kyoin-tll-ka sul-l11mek-ko topak-Ul hayessta
fA couple of believers drankand gambled.f
((Cetten) kyoin-sUli sul mekessta)((k,A) kyoin-tIll
topak hayessta)) y((kA) kyoin-till myt-4--zta) 60
CHAPTER THREE:RELATIVIZATION
Introduction discussed how 3.1. In the last chapter,I have extensively numerals--:an have only definite what I called Badjectives--Q's and introductory subjects; that is, how they cannotbe predicates of I which include Q's sentences (abbr. intro-S). Thus, in 1 and 2, ungrammatical; in 3 and 4, (B adjectives), the(b) structures are
which include A adjectives,both the (a) and(b) structures are 2 well-formed: 3 (1) a. ((certain) mendied)°((the) men weremany)
b. *((certain) men weremany)'((the) men died)
(2) a. ((etten) yeca-t11kagmohayessta)-((k1) yeca-t11
certaingirl-PL applied the girl-PL
manhassta)
were-many
b. *((etten) yeca-U1manhassta)-((kA) yeca-U1
certain girl-PL were-many the girl-EL
ingmohayessta)
applied pretty) (3) a. ((certain) girlsapplied)°((the) girls were girls applied) b. ((certain)*girls were pretty) ((the)
(4) a. ((etten) yeca-til
ertain_gliS-'EL applied
yeyppessta),
were-pretty 61
(4) b. ((etten) yeca-tilyeyppessta)-((k1) yeca-t1
certain girl-PL were-pretty the girl-PL
1.1gmohayessta)
applied
(I have parenthesized theand its Korean equivalent k.because, although I do not believe they arein the deep structure, Ibelieve their addition will facilitateinterpretation. I also parenthesize certain and its Korean equivalentetten because they areoptionally
deletable.)
In Chapter Two, I havealso discussed two rules--NR(Adj-Pr)
FORM (rule 99 in ch.2), which will be furthergeneralized in this 4 chapter, and RR-FORMATIONwhichmove in the oppositedirections.
NR (Adj Pr) FORM movesthe second conjuncts ofthe structures in
1-4 into the respectiveIntro-Sts, thus deriving thesentences of
5-8 after the applicationof other rules like RELATIVECLAUSE
REDUCTION and ADJECTIVEPREPOSING,5 whereas RR-FORMATION copies
intro-Sts into the secondconjuncts in 1-4, and derivesthe 6 respective sentences of 9-12. I will show later in thischapter
that NR (Adj Pr) FORMis a special case ofNON-RESTRICTIVE RELATIVE
CLAUSE "FORMATION.
(5) -a. -Many -men died.
b. who died were Many.
( 6) a. manhi--n yecat&l-ka -±ngmohayessta
many-REL girl-PL-Sapplied
b. *.nginoha-n etten yeca-t11-kamanhassta applied-REL certain girl-PL-S were-many 62
(7) a. Pretty girls applied.
b. Certain g,irls who applied werepretty.
(8) yeyppl-n yeca-til-ka ingmohayessta
pretty-REL girl-PL-S applied
b. Ingmoha-n etten yeca-t11-kayeyppazsta
applied-REL certain girl-PL-Swere-pretty.
(9) a. The men who died were many.
b. *The men who were manydied.
(10) a. Ingmoha-n yeca-t11-kamanhassta
applied-REL girl-PL-S were-many
b. manhl-n yeca-t&l-ka Ingmohayessta
many-REL girl-PL-S applied
(11) a. The girls who applied werepretty.
b. The girls who were prettyapplied.
(12) a. ngmoha-n yeca-t11-ka yeyppessta
applied-REL girl-PL-S were-pretty
b. yeyppi-n yeca-tAl-kaIngmohayessta
pretty-REL girl-PL-S applied
Exs. 9b and 10b are grammaticalsentencesif the relative clauses
there are non-restrictive,however, those relative clausescannot
beconceivedas restrictivewithout making the whole sentences
ungrammatical. This ungrammaticality seemsto-come- from the
impossible underlyingstructureS lb and 233:
3 11. Introductor.../..... sentence andrelate&iiiatter.s.
present analysis ofRELAT1ViZATION is based' on.-
"Thompion (1967, 1970a, -and-'1970b 63 claim that relative clauses are derivedfrom underlying conjunctions.
There are crucial differences, however,between my analysis and, theirs, which will be made clear :..s.weproceed. One of the differences is the adoption of Intro-Sin the present analysis.
An intro-S, which is introducedin Chapter TWO and briefly illustrated above (in the presentchapter), is a sentence which introduces the existence of a fact or tly.r.gin the 'possible world'
(in the sense used by G. Lakoff1971). There are two types of intro-S.Type I is an overt assertion ofthe existence of a fact or thing. In Type II, the existence ofthe item in question is not overtly asserted, but implied inthe assertion of something other than its existence.Thus, the sentences in 13 areovertly expressed in 14 (Type I), while they are merelyimplied in 15 (Type II).
Intro-S's like 13a or 13b will becalled simple intro-S.
(13) a. There was a man.
b. There is a girl.
(14) a. There was a man whotried to conquer the world.
b. There is a girl who ispractising the piano in the hall.
(15) a. A (certain) man tried to conquerthe world.
b. A (certain) girl ispractising the piano in the hall.
The synonymy of 14 and 15indicates underlying conjunctionsfor the
sentences of 15. Therefore, I propose that the sentencesok 14 are
Type II Intro-Sts vizcombinations of the sentencesin 13 and an
additional statement:
(16) a- 64
The sentences of 14 arederived from 16 byright-to-left right-to- embedding. It will be shown laterin this chapter that left embedding is a case ofNR-FORMATION (cf. rules 79and 89), except when the shared NPof the left conjunctis unspecified, neither shared NP caniloe the head noun of any relative clause unless both NPIs arecorreferential with eath other,and to be correferential they have to be[-1- specifiedj. In Chapter Two,
-Icorreferenialityl was defined as'identical specificity'.From MD's cannot be this definition, it isself-evident that unspecified RELATIVIZATION or correferential. Therefore, rules like
PRONOMINALIZATION, which arebased on thecorreferentiality of the
shared NPIls, cannot applyto any conjunction theleft conjunct of
which contains ar- specified] shared NP.
The sentences of 15 arederived from 16 byleft-to-right reference, I will name embedding (cf. rule 47).For convenience of of an intro-S and basic conjunction anyconjunction that consists (e.g. 1-4). another sentenceWhich shares an NP with the intro-S the second After the embedding inquestion has been performed, structures of conjuncts of 16 will be turnedinto the corresponding
17 below: 0 (17) a. *Man who there wastried to conquer theworld.
b. *Girl who there is ispractising the piano inthe hall- a girl Since the mere assertionof the existence of a man or cannot be without any additionalinformation about him or her the relative considered to be additionalknowledge for the hearer,
clauses in 17 are notsufficient to definitize thehead nouns, b.,.ut 24 65
-only to make the nouns rt specifiedj--thus marking the nouns with
certain, which is optionallydeletable. By the principle of
redundancy deletion, which will bediscussed in Appendix I,
relative clauses which carry onlyredundant information will be
deleted, and the respective sentencesof 15 are generated as the
result.
Type I intro-SYs (e.g. 14a and14b) are different from
Type II intro-Sys (e.g.15a and 15b) in the following manner:
Type I sentences make anovert assertion of the existence ofthe
item denoted by theShared noun. By contrast, predicates of
Type II intro-SYs whichassert the existence of the item havebeen
deleted and do not appear inthe surface. However, since Type TT
Intro-SYs must contain anexistential assertion of an item in the
deep structure--as illustratedby 15a and 15b, my referring to15
as an intro-S seemsto be justifiable. These are derived from 17a
and 17b, respectively, as aresult of the deletion of therelative
clauses. In this analysis, then, bothType I and Type II
Intro-Sts assert the existenceof a fact or a thing--the former,
by a surface assertion; thelatter, by an embedded relativeclause
which does not appear onthe surface.
Since either type of Intro-Sultimately asserts the existence
of a fact or a thing, It seemslogical to arrange the underlying
sentences so that an intro-S comesbefore other sentences whldh
assert something else thanthe existence of the itemin question
and presume such an assertion.This does not mean that all the
underlying conjuncts are sequentillyordered In 6 6
'sentence which asserts an earlierevent comes before a sentence
which asserts a later event. In other words, only theordering of
an intro-S withrespect to other conjuncts seemsto be fixed in the
deep structure. Thus 18b, which reverses thelogical ordering shown order- in 18a, is impossible, whereas19b which reverses the sequential
ing shown in 19a is stillgrammatical:
(18).a.Aman1 tried to conquer the world,but the man1 could
not rule his own home. b.*Theman.couldnotrulehisownhome,butaman.tried
to conquer the world. it (19)a""nan-failedintheexarntudce,buttheman.passed this year. b.Aman.passedtheexamthisyear,buttheman.had
failed in the exam twice.
3.12. RESTRICTIVE-CLAUSE FORMATION andthe definiteness of (1964) as NP1s. An interesting observationhas been made by Smith of determiners to the intricaterelations between her three classes
and two types ofrelative clauses-,restrictiveclauses (hereafter RR)
and non-restrictiveclauses (abbr. NR). Smith (39) summarizes hcr
observation in the followingquotation:
The three classes arenamed Unique, Specified,and Unsnecified, to indicate that they aredistinct from the traditionaldefinite and indefinite determiners:with R relatives,Unspecified determiners occur:any, all, etc.;with R and A[Appositive = myNR3 relatives, Specified:a the, g; withA relatives only, Unique: g (proper names).
n Smith's observation is illustra the-following sentences. 67
(21) *john who is from theSouth hates cold weather.
(22) *Any book, which is aboutlinguistics, is interesting.
(23) Any book which is aboutlinguistics is interesting.
(24) He pointed to acertain dog, which was lookingat him
hopefully.
(25) He pointed to acertaindog which was looking athim
hopefully.
(26) The book, which is aboutlinguistics, is interesting.
(27) The man who fixed theradio left this note.
Thus, with unique nouns,relative clauses must be NR(e.g. 20) if the sentence is to begrammatical (cf. 20 and 21) . With unspecified nouns, only a RR can co-occur(cf. exs. 22 and 23) .With
specified nouns, both RR and NR canco-occur (cf. 24 with25 and 26
with 27).
Interesting as it is, the aboveobservation seems to miss
some crucial points: (1) As to the inherent correlationbetween
the specificity and thedefiniteness of :EV's, I have madeit clear
in Chapter Two thatdefiniteness necessarily impliesspecificity,
although the reverse is nottrue. It also appears that the
uniqueness of NP/s implies theirdefiniteness. (2) The cause-effect
relaticn between determinersand relative clauses; inother words, co- what should be investigatedis not just idnich determiner can .
occur with what typeof relative clause butrather whether
specificity or definitenessis caused by the co-occurringrelative specificity of clause or by somethingelse.Thus in 25 above the a2ais caused by therelative clause, whereasamin 24 is already 68
Man in 27 hasbecome specific before therelative clauseis added. there, while the sourceof the definitized by therelative clause clause. definiteness of bookin 26 is otherthan the relative explicitly define the With these pointsin mind, I will clause is restrictiveness of arelative clause. A relative This definitionof restrictive only ifit makes an NPspecific. (1964: 91), who wrote: restextiveness isconfirmed by Jespersen The Adjuncts may be eitherrestrictive ornon-restrictive. necessarydetermination to itsprimary, former kind gives a other things which it specifies soas to keepit distinct from e.g. a red roseas distinctfrom or beingshaving the same name; a white rose.Dtphasis mine) any relativeclause that Since definitenessimplies specificity, And all other definitizes an NPmodified by it isalso restrictive.
relative clauses arenon-restrictive. ungrammaticality of21 Given the aboveanalysis, then, the that definitization is not at allsurprising:Since we stipulate is already used asdefinite cannot be performedtwice, John, which function is to noun in 21,cannot be modifiedby a RR, whose 7 of 22 seems tobe definitize its head naun. The ungrammaticality all books areinteresting conditioned bypresuppositions that not linguistics, ratherthan by the and that not allbooks are about For a NR can inherent nature ofanzas anunspecified modifier.
follow an *unspecified*NP as in 28: necessarily be consciousof freedom, (28) Any American,who would
would like theidea. that Smith is wrongin claiming It is not mycontention, however, by non-restrictiverelative that unspecifiedNPIscannot be modified 6 9 clauses; but rather that my: used in the senseof all--that is, as universal Q--is not r.specified) but E* definite].This claim will be justified later in thiz chapter.
To put differently the points made so far,if a head noun itself is already definite or specific, the relativeclause modifying the head noun will be understood as a NR;otherwise, as a RR.Thus, the underlined NP's in 29-31 are definitebefore relative clauses 8 are attached. Consequently, the relative clauses arenecessarily
NRts:
(29) yenge-lil hal-cul-lil thongyekkwan-ka
English-0 speak-NOM-0 know-REL Kim-T interpreter-S
toyessta
became
'Rim, who could speak English, became aninterpreter.'
(30) ene-111 kaci-n inkan-nln hatIngtongmul-kwa poncilcekilo
language-0 has-REL man-T lower-animals-with essentially
tallta
differ
'Man, who has language, differsessentially from lower 9 animals.'
(31) thayyangkye-4y cungsim-i-thayyang-to vmcikikoissta
solar-systemof center-as-REZisvne-tooin-fact zs-moving
*The sun which is the centerof the solar syStem, actually
moves.'
However, the definiteness ofthe underlin.ed abcve is rtfied 70
Compare the following "NPIs is lost because ofmulti-reference.
sentences with sentences29-31: celm.4n ( 29 * ) enehakca.-7i-kal Kim-T old-aEL Kim-T linguist-is-and young-;REL
sahoyhakca-ita
sociologist-is
*The Kim who is oldis a linguist, andthe Kim who is
voung is asociologist.*
(30 isipseyki-ey sal-n&n ink -nki tonglcul-ey sal-nknS
twentieth-century-In lives-RELman,-T cave-inlivesREL man-
pota te thalalchayessta
an morecorrupt-is corrupt than man of an of thetwentieth century is more
period.* kathl-n ecey ci-n thayyang-kw neyil tt1-1 thayyang-ka sun-and tomorrowrise-REL sun-S same=REL yesterday set-REL . . . - -
thayyang..-1-m-1anu-ka,lysirithalya
sun-is-MN-1D who -S wil7 -doubt
*Who would doubt.that the" SianwhiCh1 set. yesterday and -the
sun whichwill rise tomorrow.are the, same sun?* made In_ 29, Kim is no,longer definI,tetsince the reference 47+ discourse and thespeaker does by it is not uniquein the present - flarü to whom eofi an 10,0i7151,-, no:.t P4*stTloosethe -,hea;t1i,
he is referring, until.hea
purpose.As, o
existence ..c4f , enericiman 71
The same history, in which casethe NP loses itsdefiniteness. argument can be made aboutthe sun in 31*. In other words, until in the existence of aparticular item denotedby any NP is registered definite in all situations. the hearer's knowledge, noNP is inherently NI:elsare To be more exact,NPts can be useddefinitely, but no is understood as inherently -definite. Besides, whether any NP definite or not seems todepend on the scopeof the hearer s knowledge. Consider the following: tongmul-ita (32) Australia-ey kangaroo-n1n cinkiha-n
Australia-in live-RELkangaroo-Trare-RELanimal-is
*Kangaroos which livein Australia, are rareanimals. yangsunhata (33) Australia-1y namccok-eysal-n1n kangaroo-n1n gentle Australia-of South-at live-REL kangaroo-T of Australia *The kangaroos whichlive in the southernpart
are gentle. clause in 32 The non-restrictiveinterpretation of the relative found only in from the hearer sknowledge that kangaroos are understood to live onlyin the Australia. If kangaroos were relative clause of southern part of Australiaby the hearer, the non-restrictively. in normal cases, wouldalsobe interpreted in. 32 however, the same NPkangaroo -would be:taken to be definite -live not only inthe South and indefinite in33, since kangaroos In other -Words but all over, Australia--andonly inAustralia.
the -definiteness of, anVP SeemS--Ecidepend upon hearers*
unii-lue presuppositioni:1if al* NP isurideritood'-'have a etc.-=it is reference--propernoun,representatiVe'gthientc,- 72
10 definite according to thedefinition of definitenessjust suggested; otherwise, an NP has to be modifiedby a RR to becomedefinite:
Now, let us consider theunderlying structure of a RR.
Type II intro-S's play acrucial role in the accountof relativization, With respect to the as in that ofquantification in. Chapter Two. first derivation of 15 I have argued that17 is derived after the 11 conjunct of the sentencesin 16 is embedded Intothe second conjunct. specific and afterwards The relative clausesin 17 make the head nouns of 15--Type,II get deleted, allowingthe respective structures specific in intro-Sls--to be generated.Therefore, all NP's are only in basic Type II intro-Sts.RR-FORMATION is possible is an intro-S. conjunctions--conjunctions inwhich the first conjunct recapitulate those Before I present my ownanalysisI shall briefly made by Thompson(1970a and 1970b) and Drubig(1968). believe in the Thompson (1970b: 43-44)does not appear to 12 intro-S. basic ordering of conjunctslet alonethe notion of an _
To her, both 35aand 35b are equallygood as the underlying
structure of 34.And she claims that aspeaker s presuppositions conjunct to embed into about a hearer'sknowledge determine which
which conjunct.
(34) I met the girl whospeaks Basque.
(35) a. (I met girl.)-(girl. speaks Basque)
met girl. b (Girl. spea)cs Basque).(I i) conjuncts are presupposedto Thus, Thompsonclaims that when both -sentences be new information,the following four synonymous
derived: 73
(36) I met a girl,and she speaksBasque. and I met her. (37) There's a girlwho speaks Basque,
(38) I met a girl whospeaks Basque.
(39) A girl I metspeaks Basque. not the secondconjuncts, were If the firstconjuncts in 35, but sentences presupposed to beshared by the hearer,then the respective ck 40 would be derived:
(40) a. Thegirl I met speaks Basque.
b. I met thegirl who speaks Basque. presupposed to beshared by the In other words,conjuncts which are new information;they hearer are embeddedinto conjuncts which carry then definitizethe head noun. of the no-basic-order However, thefollowing consequences ordering of underlyingconjuncts hypothesis seem tosuggest that the approach, 35a and 35b arein fact is fixed: (1)Under the no-order should be semantically(if not the same.Therefore, sentence 36 with 39--since the presuppositionally) identicalwith 37, and 38, of the identicaltransformations pairs ofsentences are the outputs underlying structure. Utmany. applied to thesupposedly identical have expresseddOubt that the .native speakers Ihave checked -with
sentences aresemantically identical. (2)
difficulty arisesin this in whichone conjunctincludes a Qpredicate. 74
derived from the be synonymous with oneanother, since they are identical underlyingstructure 41. many) (41) (Certain girlsiapplied) -( (the) girlsi were
(42) Many girls applied.
(43) *Girls who applied weremany. applied, and they were (44) There were(certain) girls that
many. and they applied. (45) *There were certaingirls that were many, of many (a 0.) is not But 43 is ungrammatical,because the subject 13 ungrammatical, since manycannot be the definite. Ex. 45 is
predicate of a RR.
(46) *I met the girlsthat ware many. basic ordering forthis analysis, (3) Furthermore,since there is no of 41 would be a situationis conceivable wherethe second conjunct Notice that the presupposed in which case46 would be derived. in 40a and 40b. relative clause in46 is claimed tobe restrictivegas More exactly, arelative clause But agai 46 is ungrammatical. because of the nature o with a Q predicatecannot be restrictive not arise in afixed-order such predicates. These problems do not allowed to analysis having therestriction that aQ Is
predicate of an (4) Above all,the no-order analysis explanation of ..the makes it impossibleto give a syntactic _the analysis distinction between RRand NR. It will be shown-in can-be suggested-by the present writerthat presuppositions
syntactically explainedin a non-ad-hoc manner.
certain arid the, asdeterminers no-order analysis,the choice between 75
0. cannot of a noun seems to dependentirely on presuppositions; one
account for the choice interms of syntactic rule. The fixed-order
analysis accounts for thechoice in the syntactic ruleformulation,
as will be shownbelow (compare rule 58with 79 or 89).
Drubig (1968) recognizesthe fixed order ofunderlying
conjuncts, although he is notclear as to why theconjuncts have to
be ordered the way heclaims. He (34) graphicallyshows had
RR-FORMATION (he calls thisLEFT SISTER ADJUNCTION,since the left shown in conjunct is adjoined to theright conjunct) is formed as
47 below:
(47) LEFT SISTER ADJUNCTION(RESTRICTIVE)
suggested by Drubig willbe The formulation ofRR-FORM which is
illustrated with ex. 48below:
(48)'.
We ust-saw a ave me 48 Is sister- The structure49 is derived afterthe left conjunct of LEFT SISTER adjoined to the shared nounin the 7 6
(49)
We jussaw a jani or a jani tors'gavemehiskey
we just saw ajanitor'
Drubig (36) argues that bothNP/s have feature [-mentioned], but
LEFT SISTER ADJUNCTIONchanges the feature valuein the NP in the mentioned] is right conjunct to[4-mentioned].The featune
responsible for definitizationof the NP. As a resust, 49 becomes
50:
(50) We just saw a janitor,(and) the janitor that we gust saw
gave me this key.
The parenthesized andis derived at a laterstage, beingconditioned topics by various factors suchacasserted,- presupposed, or implied
which are shared betweentheconjuncts, andthe semantic
relatability of theconjuncts,etc. (For a detailed discussion, that 'either the cf. R. Lakoff 1970).Drubig (38) further argues
antecedent sentence Stinmy termsan introS]or the
restrictive relative clause S/ can be deleted/. Thus if the
intro-S is deleted, ex. 51will be derived. -7.. ., -
(51) The janitor that we-just saw gaveme this key.
(52) We just saw a janitor,*(_and). thejanitor.:gave
However- Drubigdoes not:mentiOn. what is-'-the fa ortlat.. determines ' which sentence is to .be tO:yield 511., or deleted--whether...
a RRto derive a structure li]ce52.One inigiat arguethat the
speaker's presuppositions about e hear er s
,. , , .
: 77
-the crucial factor. If the speaker presupposesthat the information
carried by an intro-S isalready shared by the hearer,then the
intro-S (viz, the first conjunctof 50) gets deleted. Otherwise,
the relative clause isdeleted.
But this kind of argumentis clearly contradictory.For if
object to be -^ a speaker presupposesthe existence of any fact or be needed known to the hearer, an intro-Swould, by definition, not
from the outset. Thus, hardly anyonewould claim that the (a)
sentences In 53-55 come fromthe respective (b) treesbelow:.
(53) a. The sun rises in themorning.
b.
4.
(54) a. The roof is leaking.
b. 78
fact or object is not necessary to assumethat the existence of any shared by the hearer,in order that anintro-S may be used. be conditioned by the Therefore, the deletionof an intro-S cannot presupposition that theinformation carried by anintro-S has had already been already been shared--since,if the information shared the intro-Swould not have beenneeded from the outset. is Besides the fact thatthe information isshared means that what being deleted is actuallynot an intro-S. are never As a matter of fact,I claim that intro-SIs and only under certainconditions. deleted. Only RR's are deletable between Sand S is a It seems to methat the relative distance 1 2 the speaker presupposes determining factor. In other words when apart from each that the two sentences(S1 and S2 ) are too far remembers the other for him to expectthat the hearer still repeat the intro-S, existence of the particularitem, the speaker must relatively viz, as a RR. On theother hand if thetwo sentences are copied into close to each'.other, _the RR which isSi anybody follow an automatically deleted.Only very seldom would overtly contains intro-S immediately withanother sentence which
the intro-S as a RR.Thissituation is found in 56: soon (56) John bought a car,and the, carwhichJohn bought teal
break down.
In-:a Situation-wherethe:.-sPakerflas Incorrectly presupposed
-12R, -a hearer_most thel'hearerts. knowledge-aria;aeccor,_dingy..':deted.,...the 79
be to such a Notice what the speaker'snonnal response would something like 'The carwhich John question.Would it not be boughtI where a RR is recovered? generalization: There is an apparentcounterexample to this of a RR is notallowed even there is a situationwhere the deletion item is clearlypresupposed when the existenceof that particular question in dialogue to be rememberedby the hearer.Note that B's which is used when onewants 57 below uses whichrather than what: what is used when to definitizewhat is alreadyspecific, whereas definitize what isnot yet specific. one wants toeither specify or in the last sentencein A is B's questionimplies that the RR
incorrectly deleted. It must be present. bought a Chevy.But the (57) A: John bought aFord and Mary
car will soonbreak down.
B: Which car do youmean? (1) there is no The points made sofar are thefollowing: only when it isnot (=S) (2) a RR is deleted deletion of intro-S s 1 demonstrative) and whenthe speaker used deictically(viz, as a of the head noun; presupposes thehearer to rememberthe existence when the hearer's (3) when a RR is useddeicticallythat is, deletion of that RR question contains whichrather than whatthe 80
In other words,if the is possible onlyin basicconjunction. modify it.Now, Shared NP is already definite,then a RR cannot
RR-FORM is formulatedas 58: FORMATION (abbr.RR-FORM) (58) KoreanRESTRICTIVE-RELATIVE-Q.AUSE
NPi I Z) ; NPi intro-S(Y 6 1 2 3 4 5
6) A. (1, 2, 3)-(4, (1, 3+REL)s#5,
(1, 3+REL)s#5, 6) .
FORMATION (581) EnglishRESTRICTIVE-RELATIVE-CLAUSE NP Z) (W2 NPi, X)Intro.s-(Y, E.definite] 5 1 2 3 3)), 6) 3)-(4, 5 # (REL 2)#(1, rg-definite]
2)#(1, 3)), 6) (4, 5 # (REL r-t-definite] of its prior occurrence. where 2 hasalready beenpronominalized because and Chomsky-adjoins RR-FORM (A) is acopying rule:it simply copies
the'-fc>11owing cony, an intro-Sto the sharedNP in . - which I suggest asthe underlying illustrate thisrule with 60,
. . structure of 59.RR-FORM (A)will derive 61 from looking at.him. (59) John pointedto the dog which was JohU)-"(JOhn.pointed to (60) ((Certain)dog waslociking S 81n
V NP NP IVIIIIIMIIIrz=2... certain dogs was.---oo,. -rig atJohn Johnpo- teain to NP Sti
dog.:
thatwas looking at John
at John) (Johnpointed todogs (61) ((certain)dogswas looking
that was lookingat him) asQUANTIFICATION and I assume that,such deep-level rules many languages, RELATInzATION aresimilar intheir formulation among minor ways by although theirsimilarity might havebeen blurred in In formulatingrules, I have some superficialgrammar change. of these rules.Thus tried to show themaxinuam universality that a shared NPin a RR-PORM (A) isformulated in such a. way rule that copiesthe intro-S copied intro-S is_deleted by the same of being changedto a Twh-word into the followirsgconjunct, instead The rUle and moved to thefront of therelative sentence. . . (=REL) is formulation also sugge.ststhat some sortof partic3.e iiimts that English attached to therelative clause.We have some embedded RR ratlier might have had that as'apartidle at-tached_ to n
than.as.a. relative
for the plausgolli. but in Middle The wh-pronouns wereoriginally interrogative, English which,- whom,.and -who came-into useas relative- the orderindicated) and have sincethenbeen pronouns (in in the last gaining ground at the- expenseof -that, chiefly pretentious kinds ofliterature. few centuries andin the more probably that One of the reasonsfor this preference was scholars of thecorresponding these pronounsreminded classical Latin pronouns. Rule 58' may haveohave some marked inEnglish contemporary Englishrule4 Since defin:_te NP s are noun thisrule has to be
the. further revised sothat the head noungets marked with contemporary Korean. (A), as is formulatedin 58, is valid for definite NP t.,s are not revision is necessarybecause in Korean turned into 63 byRR-FORM (A) marked. Thus, ex 62 is REFLEXIVIZATION derive.the -Otherrules like PARTICLE-INSERTIONand 16' surface structure64. kay. kalikhiessta) (62) ((etten) kayi.John palapoko-issessta).(John. John dog pointed certain dog Johnwas-looking-at 83
(63) ((etten) kay. John palapokoissessta)-(John(John
certain dog John was-looking-at John John
palapokoissess-nln)#kay kalikhiessta)
was-looking-at-RELdog pointed-to
(64) (etten) kay-ka John-Alpalapokoissessta.John-nin
certain dog-S John-0 was-looking-at John-T
caki-lil palapokoiss-te n kalikhiessta
himself-0 was-looking-at-RELdog-0pointed-to
'A certain dog was lookingat John. Johnpointed to the
dog that was looking at him.'
If the speaker presupposesthe two conjuncts of64 to be too far apart for him toexpect the hearer to rememberthe first conjunct, then the relativeclause (the underlined partof the second conjunct of 64) willremain; otherwise, it will be pronominalized by kl. The former case will,result in 65a; and the
latter, 65b:
(65) a etten kay-ka palapokoissessta. John-nin
ertain dog-S John-0was-looking-at John-T
kay-111 kalikhiessta
the dog-0 pointed-to
'A certain dog waslooking at John. John pointed to
the dog.'
b etten kay-ka John-lilpalapokoissessta - John-n&n
certain dog-S John-0 waslooking-at John-T
palapokoiss-te-n kay-l11kalikhiessta
himself-0 was-Ooking-at-RET-REL dog-0 pointed-to 84
RR-FORM (B) is a choppingrule, which movesand Chomsky- 17 adjoins an intro-S to theshared NP in thefollowing conjunct. deriving 66 from 59. The process will beillustrated again by
(66) John pointed to adog that was lookingat him.
Notice that here thehead noun (=cloa)is not a repeated NP; before 66 is stated, i.e. there is nocorreferential NP(= 19.1) the second clauseof 66 as a since the intro-Shas been moved into although the intro-S RR.This is to be contrastedwith 61, where, where it was. has been copiedinto S2 the originalS/ still remains in 61, but not in66. The head noun(=skis)is a repeated noun RR-FORM (B) Therefore, it isdefinitized only in 61.By formulating indefinite NP is theoutput as 58, I claimthat any RR modifying an
of RR-FORM (B) (a choppingrule).
RR-FORM (B) willderive 68 from 67: cukilyehayessta)(ecey na (67) (enceynka hansakona na yesterday I one-time ore sailor me tried-to-kill
sakong. mannassta)
sailor met 85
sakong-1U (68) ecey na-n&nenceynka cukilyeha-n (han) tried-to-kill-REL onesailor-0 yesterday one-time me-0
mannassta
met
'Yesterday I met asailor who oncetried to kill me.' syntactically marked as Even though inKorean definite1110's are not RR (A) and a RR(B) by such, one still candistinguish between a only RR (B)tolerates testing a RR wlth 'one' orIcertain1:18 head noun:e.g. 68. han betweenrelative clause and approach suggestedin 58, To emphasize thecorrectness of the For the derivationof RR s I will compareit with Thompson's. 94) suggests the modifying indefiniteNP's, Thompson(1967: along with her ownexamples following two stepswhich are quoted
(the numbering ismine):
(69)
eymour gave me a knitea I used a a S, into S19 Step I (optional) : anted by copying 2 original S1 Step 2 (optional): delete the applied, the resultis: - -Ifthis rule is knife which Seymourgave me. (70) I used a must is not chosen,however, the a in S - -Ifthis second option occurrencerand the be changed tothe since it is a.repeated result is: the knife that (71) Seymour gaveMe a knife, and I used Seymour gave me. 95) suggests thefollowing In contrastwith this, Thompson(1967: modifying definiteNP1s: four steps ofderivation for RR's 86
(72)
used a knifea You know that SSeymour gave me a knifa
Seymour crave me a knifea (73) I used theknife which Seymour gave me. Step I (optional): embed S2 ,into S3 bycopying since it is a Step 2 (obligatory):delete the original S2 repetition ofS.1 because this NP is a Step* 3 (obligatory): dhange a in S3 to the repeated occurrence. oft128one in Si Step 4 (obligatory): delete Si quotation is similar to The choice of Step 1andStep 2 in the first only Step 1 is similarto my my RR-FORM(B), while the choice of in these.cond RR-FORM (A). But the approachsuggested by Thompson First of all, what if we quotation above createspeculiar problems. 3, and 4 will pass over Step1, since it ioptional? Steps 2, automatically apply totheunderlyingstructure, deriving 74:
(74) I used the knife. result of° In what sense is74synonymouswith 73, .which is the having chosen optionalStep 1?SeCond, the relative ordering first between the rulechanging a to the and therule deleting the
conjunct is veryvague--in the first quotation,.Si-DELETION,,:._
precedes DEFINITIZATION;whereas_ in the second'quotation, fix the rule ordering S-DELETION followsDEFIllITIZATION-; If we 1 is-..no way...of so thatDEFINITIZATION precedesSi-DELETION,there deriving 70 underThompontsanalysis; on the otherhand; if.the . . . 73 and 74 can never.bederived:Third, . order is reversed, according to heranalysis, theidentittofthe second conjunct of 87
But my positionis "sentence 71 and sentence73 is accidental. second conjunct of71. that, on the contrary,73 is simply the between the twosentences. Note that there is nosemantic difference of S2 Last, the deepstructures are notmotivated--the occurrence of the various stepsseems is not explained--andthe optionality difference which is completely ad hoc. The presuppositional definite-indefinite distinctionof the ultimately responsiblefor the either in the deepstructure head noun is not asclearly expressed is done In my 58. or in thetransformational rules, as the derivation of 3.13. NR-FORMATION.Let us now turn to characterization of therestrictiveness of arelative a NR. From the analysis- a relativeclause is clause suggestedin the present specific--it becomesclear restrictive when itmakes the head noun cannot be definitizedby a RR that NPIs which arealready definite 27 which I repeathere for (e.g.21).21 Thus, I have shown in 26 and
convenience, is interesting. (26) The book, whichis about linguistics, this note. (27) The man thatfixed the radio left clause on the that when thehead noun isdefinite the relative if it is the relativeclause surface could beeither a RR or NR: the clause is aRR; otherwise a NR. that definitiZesthe noun, then of definitenessof book In the presentapproach, the source hearer and is notexpressed In 26 is presupposedto_be known to the 27 i8 overtlyexpressed as the on thesurface, whereasthat of man in Therefore it isnaturally expected underlined relativeclause. 88
about the head nounis only that a request forfurther information appropriate in some possible in 26. The sentence byB in 26 can be situations, but never27IB. linguistics, is interesting. (261) Ai The book,which is about
B:What book are youtalking about? this note. (27') A: The man who fixedthe radio left
?B:What man are youtalking about? structure of 75. Therefore, I suggest76 as the underlying this key. (75) I just saw thejanitor, who gave me just saw (the) ( 76) ( ( certain) j anitor 1.. .) intro-S. ( (1 this key)) janitori)( (the) janitori gave me the informationof the Intro-S In 76, the notation stands for The intro-S in76 makes that was given inthe previousdiscourse. second conjunctdefinite, the both occurrencesof janitor in the
process isillustrated by thetree of 77.
(77)
gave me this saw janito anitor. S us key
cerain janitori- . certain ,3anitori - and in both minorconjuncts (viz, The distributionof the intro-S (viz.,. S2) can beexplained by the S4) in the secondmajor conjunct is principle of distribution.TheprinciiIe of distribution The S of diagramatically shownhere: X(Y
transformed into 78after S has beencopied into bo 89
deleted in S3and S4 by the copiedRRIs are RR-FORM (A),and after
RR-DELETION. this key) -(the janitor1gave me s (78) (I just sawthejanitor.)S1 2 of 75 from78 is responsible forthe derivation The rulewhich is
formulated as79. RELATIVE CLAUSEFORMATION (79) EnglishNON-RESTRICTIVE (abbr. NR-FORN) (104NP.,Y) NP r definitej 5 6 1 2 3 4 3, q (REL + 5)4,(4, V, 6), 0, >1, 2, replacing boundary.The rule whereSt. is aphonological-phrase rather lowlevel here, sinceit is a REL + NPby a wh-wordis ignored English, thisrule also rule. Forcontemporary English-particular that RELis realized restricted insuch a way needs to befurther the head nounas such afeature on differentlyaccording to deleted NP(item also agreeWith the El personaii.REL should 80b and 80c are marking.The sentences 79) withrespect to case violate theserestrictions: ungrammaticalbecause they who gave methis key. (80) a. Ijust sawthe janitor, which gave methis key. *I just sawthe janitor, whom gave methis key- *I just sawthe janitor, formulation of79 claims that aremade in the I willemphasize the embedding isperformed in with 58.First by comparingthis rule 90
conjunct is embeddedinto the 'the opposite directionhere--the second embeds the firstconjunct into the first conjunct in79, while 58 is a choppingrule--NR-FORM does second conjunct. Second NR-FORM first conjunct, but not only copythe secondconjunct into the deletion of a NRis not completely movesit. Therefore, the when recovery ispossible.Third, allowed; deletionis allowed only conjunct should bedefinite in 79. the shared NPin the second as towhether the firstconjunct in Last, thereis no specification in 58 the firstconjunct has 79 is an Intro-S ornon-intro-S, while example of rule79 in which the to be an intro-S.Sentence 78 is an For an exampleof 79 in which first conjunct isnot an intro-S. consider thefollowing: the first conjunctis an intro-S, janitori)-(janitori gave me (81) a. (I just saw acertain
this key) janitor that Ijust b. (I just saw acertain janitor)-(the
saw gave methis key) janitor)-(the janitor gaveme c. (I just sawa certain
this key) of 79, RR-FORM(A) has to To satisfy thestructural description 81a into thesecond conjunct. apply to copy thefirst conjunct of as shownin 81b. Since an Then the head nounwill be definitized, conjunct, theif...elative intro S immediatelyprecedes the second deleted.As a clause in thesecond conjunct of81b will be operate to derive82 from result, 81c will bederived Then, 79 will
81c. key. (82) I just sawa janitor,who gave me 91
between 75 and82 is In this approach,the difference deep of a thirdconjunct in the correctly shown tobe the presence words, with75,the speaker structure of 75only. In other before knows about thejanitor, even presupposesthat the hearer NP in he does not. Since the shared 75 is stated;while with 82 since RR be definitein 79, and the secondconiunct has to necessarily hasto precedeNR-FORM. definitizes NP's,RR-FORM following rules is wellillustrated in the This orderingof the two 83 and 38, which arerepeated here as examples fromDrubig 1968:
84: lady, whonevertheless managed (83) The man whomurdered the old day. to escape, wascaught the next managed to escape,who murdered (84) *The man,who nevertheless day. the old lady wascaught the next the head nounis indefinite Sentence 84 isungrammatical, since RR; the rule later getsdefinitized by a when a NR isattached, and
ordering isreversed. analysis once morewith Let us comparethe present underlying 92) claims that85 is the Thompson's.Thompson (1967:
structure of75:
(85)
derivation of a following threesteps of And shesuggests the 92
Step 1 (optional): embed S3 into S2
I just saw ajanitor S
wno gave me tm.sIcey by the since itis a Step 2(obligatory): replace a in S2 repeated occurrenceof the NP in S1 Step 3 (optional):delete S1
I just saw thejanitor
wrio gave me tilis ey
of NR: (1) It I see twodifficulties inThompson's analysis of deleting SI inthe third step is not clearwhat the consequences information carriedby Sis rendered would be --inother words, the 1 theory being advocatedby me in this irrecoverable.Deletion, in the condition ofrecoverability. writingvis possibleonly under the that Si is notrecoverable Thompson (1967: 92) seems to believe in this chapterthat RR anyhow.But I have shownat various places ask 'What (NP) ?* to is alwaysrecoverable. One simply has to Besides, if Swere notrecoverable, what recoverthe.deleted RR. optional step-3 of would the outputbe like when wedid not choose precedes step 2 inthe the abovequotation? (2) Since step 1 in Sof 85 is not quoted approachof Thompson's,the shared NP 3 embedding of Sinto SIs performed; inother definite when the 3 2 in. 79 is absent. This umcds, the featurespecification. of item 5 generalizedS-ingle rule toderive makes itimpossible to have a 9 3
It the respective(b) structures. the (a) sentencesof 86-88 from deletable in 87,but not in 86 also has tobe specifiedthat S1 is or 88: who gave methis key. (86) a. I justmet a janitor, janitor)-(janitor gave methis key) b. (I just met a who gave methis key. (87) a. I just metthe janitor, )-(I just met ajanitor)-(a janitor b. a janitor 22 gave methis key) me thiskey. (88) a. I justmet Jobn, who gave gave methis key) b. (I justmet John)-(John is more complexthan 79. The Koreanversion of NR-FORM the paralleling 86a. In other words, Korean doesnot have a NR specification on theNPin additional feature Korean ruleneeds one 1+ definite], as isshown in the the firstconjunct, viz.
formulation of89: RELATIVE CLAUSEFORMATION: (89) KoreanNON-RESTRICTIVE
la -CY, NPi 2) (W, NPi. r+definite) E+ definite] 5 1 2 3 1, (4, re, 6 + REL)#2, only as 821a but never as821b: Thus, ex. 82 canbe paraphrased mannass-nintey, kl-ka (821) a. na-ninetten suw1-111pangkm met-and I-T certain janitor-0 now
i key-111na-eykey cuptita
this key-0me-to gave
/I just met ajanitor, who gave me 94
(821) b. na-nn i key-]Alna-eykey cu-n ettensuwi-111 janitor-0 I=T :this key-0 me-to gave-REL certain
pangklm mannassta
just met
just met a janitor, who gaveme this key./ I will Notice that thefirst conjunct of 89 maynot be an Intro-S. illustrate 89 with thefollowing sentences:
(90) a. Spain-e-3A1 salam-n&I
Spanish-0 speak-NOM-0 know-RELthe man=T
thongyekkwaa-ka toyyessta
Interpreter-S became
'The man, who couldspeak Spanish, became aninterpreter./ salam Spanish b. (kl salamthongyekkwan toyessta)-(ki
the man interpreter became the man Spanish
hal cul alassta)
speak-NOM knew silhehanta (91) a. taylyuk-eysecala-n John-nki sem-lil
continent-on grew up-RELJohn T island-0dislikes
'John, who grew up onthe continentdislikes the island./ calassta) b. (John semsilhehata) (John taylyuk- yse
John island dislikesJohn continent on grew-up structure 90c: The (b) sentenceof 90 is derivedfrom a still deeper 9 5
hal-cul alta) (90)c.(-..ettensalam.--((sa1am.Spain-e know certain man man Spanish speak-NOM
(saLmn. thongyekkwantoyessta))
man interpreter became
etten saiam.
salam.thongyekkwan toyess asalam. Spain-ehal-cul alassia
(90) d.
etten sa ara.
salam. thongyekkwantoyess SI Spain-e hal- alassta etten sadam. 0.0- e en sa am. 90c.- With respectto the RR-FORM (A) (58)will derive 90d from Korean a RR is derivation of 65, Iclaimed above that in deleted in Englishviz,when an pronominalized whereit would be derives Intro-S is not toofar away. Thus, RR-PRONOMINALIZATION 90b and 9lb, and as 90b from 90d. Now NR-FORM(89) will apply to are derived. the result therespective (a) sentences
Some possibleobjections suggested analysis 3.2. In this part, Iwill defend the possible objections: of RELATIVIZATIONby discussing some 105 96
direction of embeddingin NR-FORM; (1) objectionsconcerning the restrictive relativeclauses-- (2) objectionsconcerning 'stacked' analysis, sincedefinitization can this is not allowedin the present occur only once. in NR-FORM and theproblem 3.21. Direction of embedding this point is of ambiguity of aNR. One thing tobe mentioned at alternatively formulatedin such a way that the NR-FORMrule can be direction--the leftconjunct that embedding canbe done in either into the leftone--unless into the right one, orthe right conjunct Intro-S. The reason thatI have of course thefirst conjunct is an for the sake of formulated the rulethe way I havedone is .merely conjunct of 79 is an simplicity ofdescription: if the first of embeddingwill generate a Intro-S, thedifferent direction Therefore, in orderto make different kindof relativeclause. direction, we haveto describeboth NR-FORM applyin either one conjunctis an situations separately: (1) the situation where conjunct is anintro-S. Intro-S and (2)the situationwhere neither underlying conjunctsdoes not As far as I cansee, theordering of But neither conjunctis an Intro-S. make any differenceso long as a NR is alater parenthetical if the conjunctwhich is realized as below, a differencedoes arise. insertion by thespeaker az in 92 parenthetical NRis a secondconjunct. It is obviousthat such a who lives inTexas. (92) a. John saidthat he met Nary, in Texas) b. (John saidthat he metMary1)-(Maryi lives parenthetical insertion,NR-FORM When the secondconjunct is a lot 4,4,fir *,41Awanal......
97
NR-FORM moves from moves fromright to left only.Thus, 93, where left to right, is notsynonymous with92: lives in Texas. (93) a. Mary, whoJohn said he met, said he met Maryi) b. (Maryi livesin Texas)*(John to be In 94, however,both the (b) and(c) structures seem (a) sentence: equally good underlyingstructures of the continent, dislikes the (94) a. John, who has grownup on the
island.
1:4(Johniklas grown up onthe continent)-(John dislikes
the island) the c.(John.dislikes theisland)-(,Johni has grown up on
continent) chapter, is simplerfor But NR-FORM, asformulated in this parenthetical expressions the following reasons: (1) Since with and since in other only right-to-leftembedding applies, neither direction seems conjunctions which donot include anintro-S the presentanalysis withright-to- to make anysemantic difference, derivation of bothparenthetical left embeddingaccounts for the (2) Since In abasic conjunctiona and non-parentheticalNR1s. right-to-left embedding conjunction whichcontains an intro-S.only non-basic conjunctionneither is NR-FORM, andagain since in the present analysis direction of embeddingmakes any difference,
will account forboth cases. (79 and 89) offers The present ruleformulation of NR-FORM ambiguity problemillustrated in the an interestingsolution to the 98 following sentence fromMcCawley 1967,quoted here as 95: at 219 Main St. (95) John said that he sawthe woman who lives
95 can be paraphrasedeither as 96 or as97: at 219 Main St.1 (96) John said, 1I sawthe woman who lives that lives at (97) John said, 1I sawXI and X is the woman
219 Main'St.1
(where X can stand for any noun,such as Mary or Samts the following married sister).McCawley (17-18) suggests and 97, which.are underlying structures forboth readings of 96 quoted here as 961 and971, respectively.
(960
Prop NP-descr:X1 X said S X is called John
Prop NP-descr:X2 X2 is the womanwho lives I( X: saw X at 219 Main St 1 2 (971)
NP-descrOC NP-descr:X Prop 1 2 Xis the Xi is called John 2 woman wholives Xi said Si at 219 Main St.
Xi saw X2
of X2 is made bythe speaker of Si In other words,the description speaker of S . (= John) in 961,while in 97* it is made by the
Og 99
which is exactlyparallel, An instanceis found in Korean system of Korean(which is except that thecomplicated honorific between the speakerand the - conditioned by thesocial relations obviates the ambiguityin some situations. referent) syntactically of the subjectin 98 and that Thus, as long asthe social status markedly different,there will of the speakerof the sentence are 23 the followingsenten e: be no ambiguityin interpreting 1 sal-nln acumma-kaoass-te-lako (98) Suk-ka(na-eykey) Seoul-ey lady-S came-RET-QT Sook-Sme-to Seoul-in lives-REL
malhayessta
said lives in Seouldropped bylt 1Sook said thatthe lady who house-maid, Kyungsukis the Suppose a situationwhere Sook is a supposed to use an daughter of thehouse to whomthe maid is 'lady', and thespeaker is Kyungsukts honorific titlelike acumma In such asituation, brother, who willnever callher facummat. information of the there is noambiguity as tothe source of McCawley1s analysis,981 will underlined partin 98.According to 24 underlie 98:
(98') NP-descr:X Prop 1 Xis called Sook / 1 X-ka S-lako maihayessta 1 1 S -QT said) (X1-S 1
Prop NP-descrtX2 sal -nin acumma 2is Seoul -ey (Seoul -in lives-REL lady) X-ka oassta 2 -S came) 2 1 n 10 0
expected to call Kyengsuk -However, if the speakeris the maid who is to use that acumma and thesubject is somebodywho is not supposed
title, then 98" willunderlie 98:
(98")
NP-descr:X Prop NP-descr:X1 2 Xis called Sook Xis Seoul-ey sal-nn 1 2 X-kaS -lako malha:essta acumma 1 11 Prop
X-ka oassta 2 Now, let us see howthe present analysisof NR-;FORM explains According to the presentanalysis, the ambiguities of95 and 98. 97-reading of 95, 99a and 99b underliethe 96-reading and
respectively: Main St.) -(John (99).a..((CertairOTdoraan.Idve at 219 Intro-S
) said (John saw woman.i) -((John b.((Certain)woman.live at219 Main St.)Intro-S said(JohnsawX))-(Xiswoman.)) form of the second Sentence 95 in its96-reading is the surface there. conjunct of 99a afterRR-FORM (A) hascopied the intro-S be'realized as 99bi after the The second conjunctof 99b will
intro-S is copied byRR-FORM (A): the woman thatlives (99) b. i. (John said (John saw X))-(X is
at 219 Main St.) who is the II. John said thathe sawX (e.g. Mary
woman that livesat 219 Main St. 1 1 0 101
the application of will derive 99biifrom 99bi, after .NR-FORM (79) replaces John with he. such a rule asPRONOMINALIZATION, which on thespeaker's presupposition The deletion ofX seems to depend If the speakerpresupposes of the hearer'sfamiliarity with X: X, then hewill delete X.In that that the heareris familiar with This will generate95 case, thedeletion of who isis automatic. presupposes thatthe hearer is in its 97-reading,if the speaker leave X (e.g.100): not famiLiarwithXIthen he will (who is) the womanthat lives (100) John said thathe saw Mary,
at 219 Main St. parallels this. Conjunctions The derivationof 98 exactly 98-reading of 98, 101a and 101bunderlie the981-reading and
respectively: salta)-(Suk (acumma . (101)a.((etten)acturana.Seoul-ey Sook lady certain lady Seoul-in lives
oassta) -lakomaihayessta)
came-QT said salta)-((Suk CKeassta)-lako b.((etten)aculimla.Seoul-ey
maihayessta))0(X acummalita)) output of thesecond Sentence 98 inits 981-readingis the surface (A) has copied theIntro-S there. conjunct of 101aafter RRFORM in its 98-readIng,the secondconjunct As to thederivation of 98 after the Intro-Sis copied. of 101b will berealized as 10ibi 101bil from 101bi: NR-FORM(89) will derive
1 I 1 102
(101) b. i. Suk-ka X-kaoass-te-lako malhayess-nintey, X-nin
Sook-S X-Scame-RET-gTsaid-and X-T
Seoul-ey acumma-ita
Seoul-in lives-REL lady-is
Suk-ka Seoul-ey acumma-i-n X-ka oass-te-lako
Sook-S Seoul-in lives-RELlady-is-REL X-S came-RET-QT
malhayessta
said
If the speaker presupposesthe hearer to know that X isSeoul-ev sal-nln acumma 'the lady fromSeoul', then -i-n X IX, who is' is deleted.As the result, 98 isderived; otherwise, X will remain, as is illustrated in 102a and102b.Those two sentences are synonymous with each other, since 102bis derived from 102a afterRELATIVE-
REDUCTION has been applied:
(102) a. Suk-ka Seoul-eysal-nIn acumma-i-n Kyengsuk-kaoass-te-
lako malhayessta
ISook said that Kyengsuk, whois the lady from Seoul,
dropped by.'
b. Suk-ka Seoul-ey sal-nAnKyengsukacumma-ka oass-te-
lako malhayessta
ISook said that Lady Kyengsuk,who lives in Seoul,
dropped by.' explained in the Thus, the ambiguity ofMcCawleyts exampae is neatly
present approach by purelysyntactic rules, all of which are
natural and independentlymotivated in the grammar! 103
3.22. Stacked restrictive relative clauses. Another possible argument which could be raised againstthe present analysis of RELATIVIZATION concerns a 'stacked'clause. A stacked clause is defined by the UCLA SyntaxStudy Group (1968: 466) in 25 the following way:
Relative clauses are said to be stackedif a structure exists such that the first clause modifies thehead noun, the second modifies the head noun as already modifiedby the first clause, the third modifies the head noun asalready modified by the first clause as in turnmodified by the second clause, and so on.
When the head noun is modified by only oneRR and one or more NR's, there is no problem.Thus, 103 has only one RR and 104 has noRR,
as shown by the phrases:
(103) The short happy life ofFrancis Macomber.. That part of
his life which was hz,ppy, which wasshort.
(104) She has a short, blue, cashmerecoat. = She has a coat,
which is short, (and which is) blue,and (which is)
cashmere.
However, when two or more RR's arestacked modifiers of a head noun--
as in 105it appearsthat we have a direct contradiction tothe
generalization that a NP can be modifiedby a RR only once: this 26 is because an NP can bespecified or definitized only once.
(105) I went to buy a watchthat keeps good time that'scheaD.
(106) There is a watCh that keeps goodtime and is cheap, and
I want to buy it.
(107) There are watches that keepgood time, and some(out)of
them are cheap. I want to buy one of them. 113 104
27 In other words, But 105 is synonymouswith 107, not with106.
suggest 108 as theunderlying structure of105.
are cheap)-(somewatchesout of (108) ((Certain)watchesx of watcheskeep good time)-(Iwant to buy one out x watches ) that the first In a situation wherethe speaker presupposes hearer--i.e. that the hearer two conjuncts of108 are shared by the and also knows that someof knows of someparticular cheap watches will be like this: (1) the them are of goodquality--the derivation, second conjunct by RR-FORM first conjunct of 108is copied into the shared NP (A); (2) the secondconjunct, which now has the first conjunct as a RR,is definitized and includesthe copy of the
in turn copied intothe last conjunct. This generates the
intermediate structureof 109:
arecheap)-(some watchesout of (109) ((Certain) watchesx keep good time)-(I wantto buy one watch outof watchesx the watches that keepgood time out of thewatches that
are cheap.) conjunct of 109 in thepresent We are concernedwith only the last
discussion. not presuppose that In a situation, wherethe speaker does the derivation any of theconjuncts in 108 isshared by the hearer, the first conjunct will be the following:RR-FORM (B) will embed outcome of which willin turn of 108 into thesecond conjunct, the
be embedded intothe last conjunct.The intermediate structure of 114 105
110 will be generated as a result.
(110) I want to buy one watch outof the watches that keep gooa
time out of the watches that arecheap.
In both English and Korean, weneed a rule which deletes a possessive particle (English out of; Koreancungeyse) and an adjacent NP if an identical NP is present onthe opposite side from the NP to be deleted: that is, X out of X in English isrealized as X, after out of Xis deleted; in Korean, X cungeyse'among' is deleted from X cungeyse X, afterwhich the second X remains. I will call the rule which deletesof X or X cungeyse REPEATED-NP
DELETION (abbr. REPT-NPDEL).REPT-NP DEL deletes out ofthe watches
from both the last conjunct of109 and that of 110.Then 105 will
be ambiguously derived.
In Korean, we have moreinteresting evidence in favor of
the present analysis. If the first conjunct of 111is embedded as
a RR into thesecond conjunct, 112 will bederived:
(111) ((etten) sikey-tx sikan calmacninta).((lW sikyey-tllx
certain watch-PL time verycorrect the watch-PL
cungeyse (yakkan-ly)sikyey-tAlssassta).(na (kA)
among some-of watch-PL were-cheap I the
sikyey-tilcungeyse sikyeyhanasako-siphta)
watch-PL among watch one buy-want 106
sikyey-tilcungeyse (yakkan-iy) (112) ((sikan cal mac-nin)s x time very correct-RELwatch-PLamong some-of
sikyey-tilssassta)-(na (k) sikyey-tl cungeyse
watch-PL cheap I the watch-PL among
sikyey hana sako-siphta)
watch one buy-want. into the last conjunct, If the first conjunctof 112 is embedded it stands, it needs 113 is derived. Although 113 is grammatical as adjustment rules to become awell-accepted structure. Sudh delete sayea:Lil adjustment rules includeREPT-NP DEL, which will identical cungeyse loutof the watches' beforethe occurrence of NP-cungeyse NP (= sikyey).The degree ofdeletability of Korean In other words, seems to be muchlower than that ofEnglish of NP. of in English of 1..? hasto be deleted almostobligatorily regardless identical NP; but in the distance betweenthat phrase (of-NP) and an by aa identical Korean, unless NP-cungeyseis immediately followed deleted.Thus, NP, the phrase(NP-cungeyse) is only optionally,
sentences 113-115 areall grammatical, but116 is not:
(113) na-nln ((sikan-kacal mac-nln) sikyey-tll cungeyse
(ssa-n) ) sikyey hana-lilsako-siphta SS cheap from 'I want to buy oneof the watches that are
among the watchesthat keep good time.'
(114) na-rAn ((sikan-kacal mac-nln)slkyey-t11 cungeyse
) sikyey hant-111sako-siphta. (ssa-n)SS (ssa-n) s) sikYey (115) na-nki((sikan-ka cal mac-nln) s
hana-111 sako-siphta 1 6 107
(ssa-n) ) sikyey-t4.1 (116) ?na-nn((sikan-ka cal mac-nln) SS
cungeyse sikyeyhana-111 sako-siphta but perhaps These Korean examplesillustrate a very interesting, obvlous, feature of language,namely that, just as Inconstructing for the final touch, a building, soin language, we need processes as well as the morebasic building processes. offers a very The suggested analysisof stacked RR clauses Thompson neat solution to oneof Thompson's problems(1970b: 50): realized that 117 and118 have different underlyingrepresentations, but she did not specifyin what way they aredifferent.
(117) Three boys who had beards wereat the party.
(118) Three boys, who hadbeards, were at the party. these, keeping in Let us revert to theunderlying structures of First of all, if one mind the suggestions madein this chapter.
-remembers the directionof embedding in RR-FORMand NR-FORM, viz. conjuncts, while NR-FORM RR-FORM embedsleft conjuncts in right conjuncts, it is easy to moves rightconjuncts in-the left (1) (boys had reconstruct the orderof the underlyingconjuncts:
beards) and (2) (boys wereat the party.j, Since 117 has
conjunct (1) as a RR,and since RR-FORM movesthe left conjunct into conjunct (2) in the the right conjunct,conjunct (1) must precede conjuncts are in the underlying structure of117. But in 118 the
opposite order in the deepstructure, since NRFORM movesconjuncts
whicn RR-:FORM moves them. in the oppositedirection from that in
So far as thesetwo conjuncts areconcerned both conjuncts are 108
'ordered as shown In 117aand 118a, respectively: at the party) (117) a. ((Certain) boysihad beards)-(boysi were had beards) (118) a. ((Certain) boysi wereat the party)-(boysi is not talking about any In the first conjunctof 117a, the speaker
particular group of boys whohappen to have beards,but about all
the boys in question whohad beards. Therefore, unless 117a.is
modified somehow, it will meanthat all the boys whohad beards were Since 117 means at the party. This is not the meaningof 117.
only 'some of the boyswho had beards', 117awill have to be
revised as 117b: out of boysx (117) b. ((Certain) boysxhad beards).(some bo s
were at theparty)
The third conjunct thathas to appear in thedeep structure numeral (a B of 117 and 118--viz.(boys were three) --includes a can have onlydefinite NP's adjective).Remember that B adjectives has to as subjects. So, the subject NP lamof the third conjunct In the bedefinite.ThisNPiscorreferentialwithboysunderlined in 118a. second conjunct in117b, and with boys1In the first conjunct third co,ajunct is The long and shortof all this is that this after the first posited after the secondconjunct in 117b, but
conjunct in 118a, asshown in 117c and 118b,respectively: out of boys (117) C. ((Certain) boys had beards).(some boys x
were at theparty).(boys were three)
d. (Some boys outof boys who hadbeards were a...7. the
-par-ty)- (boys were three) 109
(117) e. Three boys out of the boys whohad beards were at the
party.
(118) b. ((Certain) boysx were at theparty)-(boysx were three).
(boysx had beards)
c. ((Certain) boysx,who had beards, were at the party).
(boysx were three)
The derivation of the surfaceforms of 117 and 118 from these underlying structures is obvious. After the first conjunct of 117c is embedded into the second conjunctand definitizes the correferential NP (= basx), 117d isgenerated. Conjunction 118c
is generated after the last conjunctof 118b is embedded into the
first conjunct by NR.FORM.NR (Adj Pr) FORM, which wasdiscussed
in Chapter Two (rule 96), novesthe last conjunct of 117d and 118c
into the first conjunct there andgenerates 11,e and 118, respectively.
Sentence 117 is derived from117e, after REPT-NP DEL deletes of boys.
The semantic difference between117 and 118 is shown in the
underlying representations, viz. 117cand 118b. Also, the
derivation from these basic structuresis effected by three already
existing rules--viz. RR-FORM, NR-FORM,and REPT-NP DEL.
Thompson (1970b: 43) has also pointed out that 119a and
120a have different meanings. The semantic difference is causedby
the different underlying structuresof 119b and 120b respectively:
(119) a. I have three students who areflunking.
b. ((Certain) studentsare flunking)-(I have some
studentsof students )-(studentsare three) 11$ 110
'4; (120) a. I have three students, and they areflunking.
are three)-(students b. (I have students x)-(studentsx x are flunking)
To derive 119a, it is necessaryto let RR-FORM apply to the first two conjuncts of 119b, then tolet NR (Adj-Pr) FORM apply to this output and to the third conjunct,and finally to let REPT-NP DEL apply to the final output. If RR-FORM (A) applies to the first and last conjuncts of 120b, and isfollowed by the application of
NR (Adj-Pr) FORM to the first twoconjuncts, 120c is generated:
(120) c. I have three students, andthe students that I have
are flunking.
RR-DEL (deletion of the underlinedclause in 120c) and
PRONOMINALIZATION (changing the studentsto they) together generate
120a from 120c.
In sum, I suggest that anysentence with NP's modified both by Cats and RRIs hasthree conjuncts in its deep structure: a first conjunctwhich is ultimately realized as a RR, athird conjunct which has a Q as .J.tspredicate, and a second conjunct which has a DIP. out of NP,construction such that NP. is 1 conjunct and such that DTP-is correferential with an NE' of the last 3 correferential with one of the firstconjunct.
Other related problens
3.3. In this section, three relatedproblems--(1) the derivation of generic nouns, (2)identity and referentiality of
NPts, and (3) implications madeby the hierarchy of anaphora--
t."). 111 will be discussed. The position which the authorhas taken on reoresentative generics and correferentialitywill be further justified.28
3.31. Derivation of generic nouns. It has been previously claimed that there are two usages ofgeneric nouns:representative and partitive.A representative generic refersto:all the mer.bers of a class. A partitive generic refers tonon-specific members of some class.Compare 121 and 122:
(121) Women, who are talkative, are goodinformants.
(122) Women who are talkative are goodinformants.
(123) Talkative women are good informants.
Women in 121 is an example of arepresentative generic, and any representative generic is definite, inview of the definition of
definiteness suggested in the presentdissertation--viz, that a NP
isdefinite if the speaker presupposesits existence to be shared by hearers. Since the head noun (women) in121 is already definite, the modifying relative clause can onlybe non-restrictive. On the other
hand, women in 122 is not a.representative generic, and since itis
not specific, it cannot bedefinite.Therefore, the relative clause
has to be a RR. This is because an unspecified NPcannot be
modified by a NR.Since both'121 and 122 canbe further reduced
to 123 by RELATIVE CLAUSEREDUCTION and ADJECTIVEPREPOSING, 123 is
ambiguous. Sentence 123 has yet a thirdreading: those women who
are good informants aretalkative.The underlying structure ofthe
third reading is 124a. Sentence 123 with the thirdreading carries 112 heavy accent on thepreposed adjective (=talkative). are (124) a. ((Certain) women aregood informants)-((the) women
talkative)
b. The women who aregood informants aretalkative..
c. Talkative womenare goodinformants (= 'It is
talkative women that aregood informE-Its).
RR-FORM (A) (rule 58') Exs. 124b and 124c arederived from 124a by Note that and NR (Adj-Pr) FORM(ch. 2, rule 96),respectively. rule 96 of ChapterTwo), and RR (A)-- NR(marked NP accent; see
PR (B)EENR(unmarked accent). and 124b indicates that The semanticdissimilarity between 122 underlying structure of 124a cannot underlie122. What could be the quantifiers, which 122?Carden (1970:35) claimsthat 'generics and class and share ahigher- form a semantic class,also form a syntactic explain why this shouldbe the verb deep structure'. But he does not is clearly revealedin case.The relationbetween generics and Q's 121 and 122, whichwill be the followingunderlying structures of
justified later: talkative)). (121) a. ((Womenx aregood informants)-(womenx are
(mamenx are all) women aretalkative) b. (All women aregood informants)--(all good informants. c. All women,who are talkative, are talkative)-((womenx are good (122) a. ((Certain) womenx are
Informants).(womenx areall))
b. Namenx that aretalkative are goodinformants).
(womenx th';.c are talkative.areall) 122 U3
(122) c. All women that are talkative are good informants.
In the derivation of 121, NR (Adj-Pr) FORM embeds the lastconjunct of the (a) structure into both conjuncts within the firstmajor 29 conjunct by the principle of distribution. This generates the
(b) conjunction, from which NR-FORM (rule79) generates the (c) sentence. In the derivation of 122, the (b) conjunction is derived after the first conjunct of the (a) structure isembedded into both conjuncts in the second major conjunct by RR-FORM. NR (Adj-Pr) FORM generates 122c from 122b.
At this point, we need a rule which creates generics outof universally quantified NP1s. This rule, GENERIC FORMATION, seems
to be wide-spread among the languages ofthe world.The rule simplv
drops all in English and the correspondingphrase motri--n Tally in
Korean, cf. the followingformulation:
(125) GENERIC FORMATION: OPT W, SAll 1, NP X ==> 1, 957 3 motA-nj 1 2 3
Rule 125 derives 121 and 122 from therespective (c) sentences.
The derivations of 126d and 127d in Koreanexactly parallel
those of 121 and 122 in English:
(126) a. ((yeca-t4m1xmalssengkkuleki-ita)-(yeca-tiax girl-PL trouble-makers-are girl-M
motu-ita) sutasl1ewta))-(yeca-.U1x are-talkative girl-PL all-are 123 114
(126) b. (moti-n yeca-t1lxmalssengkkuleki-ita)-(mot&-n
all-REL girl-PLtrouble-makers-are all-REL
sutasllewta) yeca-tilx girl-PL talkative
c. sutasIlew-Inmotl-n yeca-tll-nlnmalssengkkuleki-ita
talkative-REL all-RELgirl-PL-T trouble-makers-are
d. sutasllew-th malssengkkuleki-ita
'Girls, who are talkative, aretrouble-makers.'
(127) a. ((etten)yeca--Ulx sutasllewta)((yeca-tllx
certain girl-PL talkative girl-PL
malssengkkuleki-ita)-(yeca-tIlxmotu-ita))
trouble-makers-are girl-PL all-are malssengkkuleki-ita)-(sutaslew- b. (sutasllew-Inyeca-.U1x talkative-REL girl-PLtrouble-makers-are talkative-
yeca-t1lx motu-ita)
REL girl-PLall-are
c. motl-nsutas-Ilew-ki yeca-til-n&nmalssengkkuleki-ita
all-REL talkative-RELgirl-PL-Ttrouble-makers-are
d. sutasilew-kiyeca-tll-n.&n malssengkkuleki-ita
'Girls who are talkative aretrouble-makers,' the last conjunct In the derivationof 126, NR (Adj-Pr) FORM moves within the first major of the (a) structureinto both conjuncts the (b) conjunct by the principle o istribution, thus deriving conjunction, from whichNR-FORM (rule 89) derivesthe (c) sentence. of the (a) structure In the derivationof 1277 the first conjunct
124 115 is embedded into both conjuncts in the second majorconjunct--by the principle of distribution--and derives the (b) conjunction,from which NR (Adj-Pr) FORM derives the (c) sentence. Rule 125 generates 126d and 127d from the respective(c) sentences. The underlying structure suggested for 127d explains the difference between generic NPIs modified by RR (e.g.127d) and other indefinite ad'ective + NP phrases like 128c, the deep structureof which does not include a universal quantifier: cf. 128a.
(128) a. ((etten) yeca-t1lx thgmohayessta)*(yeca-tllxsutas0.ew- certain girl-PLapplied girl-PL talkative-
essta)
were
b. kigmoha-n yeca--U1-..ka sutasIlewessta
applied-REL girl-PL-S talkative-were
c. sutas&lew-&n yeca-t11-kaIngmohayessta
*Talkative girls applied.*
In 128c, the primary accent is on sutasllew-in*talkative*, since the adjective has been moved from the secondconjunct. Sentence 128b
is derived from 128a by RR-FORM, and NR(Adj-Pr) FORM generates 128c
from 128a.
So far, I have merely shown without anyjustification that
generics can be analyzed as having been derivedfrom NPIs modified by universal Ws. The negation of sentences showsthat the
approach suggested above is the correct one.When a sentence with
a preposed adjective(e.g. 124c and 128c) is within the scope of a
1 2,5 116 negative marker (= NEG), theNEG negates the heaviest-accented part--the preposed adjective.Thus, negations of 124c and 128c are given below with their paraphrases:
(124) d. I don't think thattLkativewomen are good informants.
e. It is not talkative womenthat are good informants.
(128) d. sutasilew-in yeca-til-kaIngmohayessta-ko-nin
talkative-REL girl-PL-S' applied-QT-DLM
sayngkakha-ciani-hanta3°
think-NOM
'I do not think that talkative womenapplied.'
e. klgmoha-nyeca-til-ka sutaslew-4n yeca-.U1-ke
applied-REL girl-PL-S taikative-RELgirl-PL-S
ani-iessta
not-were
'The girls who applied were nottalkative girls.'
Compare these with the negationsof generics (e.g. 122 and
127d, 127d). Exs. 1226and 127e, which arethe negations of 122 and are synonymous with122e and 127f, respectively, tut neverwith
122f and 127g.
(122) d. The claim that womenwho are talkative are good
informants is not correct.
e. It is not truethat given any woman, if sheis
talkative, then she is a goodinformant.
f. Good :male-informants are not talkative. 117
(127) e. sutasllew-In malssengkkuleki-la-nin
talkative-REL girl-PL-T trouble-makers-are-REL
mal-nin kecis-ita
claim-T false-is
'The claim that women who aretalkative c%re trouble-
makers is false.'
f. yeca-nin nuku-itenci sutasilew-imyenmalssengkkuleki-
girl-T whoever talkative-if trouble-maker
mal-nln kecis-ita
is-REL claim-T false-is
'The claim that whatever women aretalkative are
trouble-makers is false.'
g. malssengkkuleki-i-nyeca-t11-nkn sutasIlew-c! ani-hata
trouble-makers-are-REL girl-PL-T taikative-NOMnot-do
'Girls who are trouble-makers are nottalkative.'
Thus, a sentence like 127e is usedwhen the speaker's expectations based on the generalization that alltalkative women are trouble- makers has been upset:by a certain womanwho is not a trouble-maker.
From now on, I will use only122 by way of an illustration.
A logical representation of122e9 which is a paraphraseof 122d, Is
given below as I22g:
(122) g. 00 (Talkativex Good Informantx
'It is not the case that given any x,if x is Talkative,
x IsaGoodInformant.'
To reconstruct the correctunderlying structure, the following if:4;7 118
used:31 - replacement rules will be
(129) De Morgan's Theorem:
^-(p q) (.-.pv ~q)
'negation of a conjunctionof positive sentencesis
equivalent to a disjunctionof negative sentences.'
(130)_Material Implication:
p q v q)
'implication of q by pis equivalent todisjunction of
negative p and positiveq.' 122h, which is in turn According to 130, ex.122g is equivalent to
equivalent to 122i by129:
(122) h. ^,(N)(^-Taikativex v GoodInformantx)
'It is not the casethat given any x, xis either Good
Informant or notTalkative.'
--Good Informant) i. -%'(x) (Talkativex x 'It is not the casethat given any x, xis never both
Talkative and not GoodInformant.'
rule which is givenhere as 131: At this point, weneed a Q negation
(131) Quantifier,-Negation:32
(v) My S. C3t, )6,- My (-rvI5 ) Rule 131 means thatthe negation ofuniversal quantification (3v ) of a negated Is equivalent tothe existentialquantification 1221 is equivalent to122j: statement(''d1/).According to 131, ex. 119
(122) j. (Elx) (Taikativex-,-GoodInformantx)
'There is at least one x suchthat x is both Talkative
and not a GoodInformant.'
Thus, all the representationsof 122g-j are equivalent to each other.Notice that 122e is a paraphraseof 122d, which is the negation of 122--a sentence with ageneric. Since 122e is represented as 122g, the underlyingstructure of 122 (with affirmative meaning) should be the negation of 122g: anything negative logically becomes positive if negatedagain. Sentences 132g-j are the rough Englishrenderings of the negations of122g-j.
These are the underlyingstructures of 122, a sentence with a generic noun.
(132) g. For all women, it is truethat, if they are talkative,
they are good informants.
h. It is true of every womanthat she is either a good
informant or not talkative.
i. It is true of everywomanthat she cannot be both
talkative and not a_ good informant.
j. There is no woman such thatshe is both talkative and
not a good informant.
It is clear that theclaim =which the analyses of122 and 127
are based--viz,that generic nouns arederived from nouns modified
by a universalQ-has been justified in thelogical representations
and replacement rules.
3.32. Identity and referentiality_ofNP s. Identity and correferentiality of NPts frequently serveas conditioningfactors 129 120
EQUI-DELETION. Thus, 'for PRONOMINALIZATION,REFLEXIVIZATION, or
Kartunnen (1971:3) liststhe followingthree deletability
conditions, which mustobtain simultaneously:
(133) DELETABILITY CONDITIONS:
a. IdenticalConstituent Structure b. Morpheme-for-MorphemeIdentity c.Correferentiality not surface conditions, It has to beemphasized that 133a-c are at a deeplevel. The identical but conditionswhich are operative subject to deletion, even NP (viz, all theDemocrats) in 134 is not conditions--133a-c--are met onthe surface. though all the above to be elected. (134) All the Democratswant all the Democrats
(135) John wants tobe elected.
(136) John wants Tomto be elected. superficially identicaland The impossibilityof deleting the
carreferential NP in 134is clearly revealedfrom the following
deep structure of134: &dn.] be elected). (134') CDemocrats want((Democrats
(Democrats beall)))-(Democrats are all) there is nocorreferentiality In the first majorconjunct cu. 134', and either of the between the subjectof the matrix sentence condition (c) of 133is embedded conjunctsubjects; and therefore
violated. possible to Only under the presentapproach does it-seem -a neat andclear maintain these conditionsfor deletability
manner.Consider-thefollowin 121
(137) I bought a red flower, and) I want to give the flower
to Mary.
If relative clauses were underlyingembeddings, a serious difficulty
would arise with respect to theapplication of RELATIVIZATION or
PRONOMINALIZNTION to a sentence like137--133h is violated. The
NP of the first conjunet in137 is not identical with that ofthe
second conjunct in tnis example.But in the present approach--
which claims that all relative clauses arederived from independent
conjuncts in the deep structure--thedifficulty discussed above
does not arise. In what is suggested here,137 will have 138 az
its underlying structure:
(138) (I bought (certain)flowerx)-((I want to give flowerx
to MarD-(flowerx is red))
I To derive 137, the firstconjunct of 138 is copied into thesecond
conjunct by RR-FORM (A)(rule 58), and as the result, theshared NP
in the second conjunct isdefinitized (cf. 137a).Then the RR is
deleted, since the Intro-Simmediately precedes it (cf.137b);
there is no danger of thehearer!s'having forgotten the information
carried by the intro-S.Finally, NR (Adj-Pr) FORM. embeds thelast
conjunct into the first conjunctthus deriving 137 from 137b:
(137) a. (I bought (certain)flower)-(I want to give the flower
that I bought to Mary)-(floweris red)
b. (I bought (certain)flower)(I. want_ to 'give the flower
-to MerY0 (flower
Let us now consider 139: -4, ,
(139) I want to give a redflower that Ibought to Nary. 131 122
If one wants to Sentence 139 isobviously synonymouswith 137. embedding, 137 may be analyze a relativeclause as an underlying However, one cannotderive 139 the underlyingstructure of 139. conditions of 133. from 137 withoutviolating the deletability several reasons why But it is being nowproposed that there are also underlies 137. 139 can be easilyderived from 138, which 32 made in the following manner: The derivationof 139 from 138 is 138 is copied into thethird (1) After thefirst conjunct of produced is deleted.And conjunct by-RR-FORM(A), the RR thereby the second conjunctby RR-FORM the first conjunctis embedded into (2) After the secondconjunct (8). Then 140a isderived. by NR-FORM (rule79), of 140a is embeddedinto the first conjunct
140b is derived: bought to Mary,(and) (140) a. I want togive a flower that I
the flower is red. which is red, \ b. I want to give aflower that I bought,
to Mary. will derive 139 from (3) RELATIVE CLAUSEREDUCTION and ADJ-PREPOSING of the deletability 140b. In the proposedanalysis, then, none all conditions have to beviolated, since inthe deep structure conjmicts. adjectives werepredicates of Individual Sentence The same line ofargument holds forKorean too
141 is synonymouswith 142: 123
(141) na-nln (han) pu1ki7n kkoch-lilsass-nintey, ki kkoch-15i1
I-T onered-REL flower-0bought-and the flower-0
Mary-eykey cuko-siphta
Mary-to give-want
II bought a red flower, and Iwant to give the flower
to Mary.I
(142) na-nin na-ka sa-n pulk-inkkoch-lil Mary-eykey cuko-siphta
I-T I-S bought-REL red.;REL flower-0 Mary-togive-want
II want to give a red flower that Ibought to Mary.'
The sentences of 141 and 142 arenot synonymous for the reason that
142 is derived from either sentenceof 141, but because both of the
sentences share the same underlyingstructure, viz. 143:
(143) (na (etten) kkochxsassta).((na kkochx Mary-eykey
I certain flower bought I flower Mary-to
cuko-siphta)-(kkochx pulkta))
give-want flower red
For 141, the derivationproceeds as follows: (1) the first conjunct
of 143 is copied into thesecond conjunct there by RR-FORM(A), and
the RR in the secondconjunct--(directly following theIntro-S)--is
deleted; (2) the third conjunctis embedded into the first'conjunct
by NR-FORM (rule 89). Forthe derivation of 142: (1) the first
conj=ct.: of 143 is copied-intothe last conjunct by RR-FORM(A),
and the RR resulting therebyis deleted; (2) the first conjunctis
embedded into the second conjuncttry RR4'ORM (B); (3) the third
conjunct is embedded into whatiszoaow the first conjunct by Mor
- 124
3.33. Implications made by the hierarchyof anaphora.
G. Lakoff (1968b:33) hasdivided definite NP's intofour groupe
and proposed the hierarchy shown below:
(144) 1. Proper names (e.g. Dirksen) 2. Definite descriptions(e.g. the man in the bluesuit) 3. Epithets (e.g. thebastard) 4. Pronouns (e.g. he)
Lakoff claims that the samerelation between antecedent and
anaphora exists between NPtsof higher group and Milsof lower
group as the relationthat exists between definiteNP's and
pronouns.Thus, the (b) sentences below areall ungrammatical,
while the (a) sentences aregood only if the underlinedNP's are
correferential:
(145) a. Napoleon enteredthe room, and he announcedthat
Jean-Luc would hang.
b. *He entered the room,and Napoleon announced that
Jean-Luc would hang.
(146) a. Emoleon entered the room,and the bastard announced
that Jean-Luc would hang.
b. *The bastard enteredthe room, and Napoleonannounced
that Jean-Luc would hang.
(147) a. Napoleon enteredthe room, and the emperorannounced
that Jean-Luc would hang._
*The emperor entered the roomand Napoleon announced
that Jean-Luc=would, 125
(148) a. The emperor entered the room, and he announced that
Jean-Luc would hang.
b. *He entered the room, and the emperor announcedthat
Jean-Luc would hang.
(149) a. The emperor entered the room, and the bastard announced
that Jean-Luc would hang.
b. The bastard entered the room, and the emperor
announced that JeanLuc would hang.
(150) a. The bastard entered the.room, and he announced that`
Jean-Luc would hang.
b. *He entered the room, and the bastardannounced that
Jean-Luc would hang.
This indicates that there is some sort of hierarchy,which
determines the ordering of definitizing clausesin the underlying
structure. As expected definite NPfs, which are morerestricted
in the sense that the scope of theirreference is narrower, are
higher in the hierarchy.Proper nouns are the most restricted and
therefore highest in the hierarchy: consequently, they are ordered
first. The pronoun he can refer to any male and socannot occur
before a uroper noun in the deep structure.To put it differently,
if two definite .NPfs which are correferential areused in the same
discourse, the possibility of ambiguity as tothe identity of
referent is least where there is a proper nounand greatest where
there is a pronoun.To La e.ffectivel then, it isunderstandable
that the identity of any NP is confirmed with aterm Wliich has the 1 91r r 126 least possibility of ambiguity;after the identity of the two has been clearly established,effective communication will not be diminished by referring to the NP with asomewhat more ambiguous 127
CHAPTER FOUR: TOP ICALIZAT ION:
NXN VS. KA
Introduction of an NP whichhe 4.1. Geach (1950)discusses two uses which in turn were calls the formal useand the material use, to characterize borrowed from Aquinas.Geach (465) goes on these in thefollowing manner: stands for an phrase is usedmaterially if it A noun or noun if it is a object (suppositum, resnaturae)-, but formally, essentia) predicate expressamg aproperty(forma, natura, lay a.tself to name someobject If a noun or nounphrase is used be regarded asmaterial, in the environment,then its use must whether it is a propernoun or not. is used formally, Thus, the subjectNP of anexistential sentence is 1. Accordingly, since it fills theblank of 'something with the (b)sentences: the (a) sentencesbelow are synonymous
(1) a. There was adog.
b. Something was adog.
(2) a. A dogchased Mary.
b. Something was adog; and the dogchased Mary. structureone Ex. 2a hasconjoined predicatesin the deep and the otherasserting its asserting theexistence of a dog, 2a as a singlesentence does having chased Mary. In other words, dog but that of aspecific dog, not assert theexistence of any of the sentence! viz, the one thatis specified bythe predicate the existenceof this After the utteranceof this sentence, of-anyib.earers specific dog is alsohereafter apresupposition 12? 128
will not and is therefore definitized.Naturally, definite nouns is illustrated in 3: occur as subjectsof existential sentences, as
(3) *John existed.
(4) Nimrod existed (= didexist).
But then, how could itbe-that Nimrod in 4--whichis as definite
as John--couldbe used as the subjectof an existential sentence? in 31 The difference between3 and 4 is to be explained asfollows: by both the John is definite, i.e.his existence is presupposed
speaker and hearers, sothat it is not necessaryto assert his when the existence existence afterwards, whereas4 is acceptable only feels the need of Nimrod is doubted bythe hearers and the speaker of 'the mighty to convince the hearersfurther of the existence
hunter against Jehovah'.
Indefinite NP's modified byadjectives, which I have
extensively discussed in theforegoing chapterscould all be 1 We can expand considered to be instancesof the formal use ofNP. -include those 14P/s which our term here, sothat NP s used formally
supply the new information-in a sentence. In Korean, the
distinction between NP tsused formally and NP
syntactically marked byparticles. Formal NP IsIn an expanded are- never sense, may includedefinite NPfs; butindefinite NP's
used materially.A plausible solutionto the long-debated an suggests- notoriousprobleM of n.4.n vs. ka (JaPanesewa''Vs. 129
An NP usedformally DefinLth NP aridTOPICALT7AT ION. which it is the partof a sentence cannot betopicalized, since differently,definiteness delivers the keyinformation. To put it condition for an NPto be is a necessary,although notsufficient, 2 its intro-S cannever betopicalized. bopicalized. Thus, an NP in the folladingsentences: This is clearlybrought out in
(5) a.m4111-4n yeca-tIl-ka ingmohayessta
nany-REL girl-PL-Sapplied
*Many girlsapplied.*
b. ((etten)yeca-tIllingmohayessta)-((k) the girl-PL certain girl-PLapplied 3 manhassta)
were-many yeca-ti *((etten) yeca-tllimanhassta) ((kl)
&ngmohayes sta) Ingmohayessta (6) a. tases(myeng-ly) yeca-til-ka applied five person-of girl-PL-S
*Five girlsapplied ((ci) yeca-t11..tases b. ((etten) Ingmohayessta) -the- five amgirl-PLappiied
persons-were (myeng)-iessta) -((ki)yeca-t&3.1 ((etten) yeca-tilitases
Ingraohayessta impossible because numerals The (c)structures above are 130
have indefinite NP's -belong to B adjectives(cf. 2.2), which cannot the (c) sentenceshave B as subjects--thefirst conjuncts of (2.21), adjectives as predicatesof intro-Sts. In Chapter Two modified by adjectives areformally- I showed thatindefinite NP's which always negates new used NP's bypointing out that NEG, Naturally, information in a sentence,negates those adjectives.
such indefiniteNP's cannot betopicalized. manner: TOPICALIZATION isformulated in the following
TOPZN): (7) TOPICALIZATION (abbr.
NP + Particle, & definite'
1 2
(2 + nki)#:(11 0, 3) 8 are derived fromthe (a) Thus, the (b)-(d) sentences in ex.
sentences by TOPZN:
(8) a. John-kwasuyoil-eyhakkyo-eykanta4
John-with Wednesday-onschool-to go
go toschool with John=Wednesday.' 131
(8) b. John-kwa-nln,suyoil-ey hakkyo-eykanta lAs for John, I goto school withhim on Wednesday.'
NP 1-1- topic)
NP r+ coin) John kwa nn
c.suyoil-ey-n&n,John-kwa hakkyo-eykanta lAs for Wednesday,I go to schoolwith John.
NP r topic]
NP time) t cornj r+ did suyoil eynin na John kwa hakkyo ey kanta d. hakkyo-ey-nln,John-kwa suyoil-eykanta 'As for school,I go therewith John onWednesday.' .... . , ' ..
NP r+ topic
NP f+ dir3 hakkyo ey,
reStrictionS'OriTTOPZW-Vi I Further :. . The-iiisbri-ity'[.:TP modified by.adjectaVed.-.
emr;;;,;: 132
Thus, 5a - doesnot definitize a head nounwhich is in an intro-S.
and 6a will becomeungrammatical if the subjects aretopicalized:
(5') a. manh-in rigmohayessta
many-REL girl-PL-T applied
*tAs for many girls,they applied.'
(6') a. *tases (myeng-4y) ingmohayessta
five person-6f girl-PL-T applied
*tAs for five girls,they applied.' NP's in 5'a and 6ta are As expected, however,if the whole subject
definitized by RR-FCRM(cf. 31 rule 58), then theNP's can be
topicalizedg.s shown. below: siptay-iessta (5") ipeney Amgmoha-nmanh.-41-1 yeca-tIl-nln motu
this-time applied-RELmany-REL girl-PL-T allteen-agers-were
'As -For the many girlsthat applied this timethey were
ill teen-agers.'
(6") ipeney Ingmoha-n tases(myenq-iy) siptay-yessta
this-time applied-aEL five(person-of) girl-PL T all teen-
agers-were they were all 'As for the five girlsthat applied this 'time,
teen-agers.
Before proceeding further
definiteness.
definiteness as something that there a#eseveral sPeaker and hearers.I have also claimed, 133
definitization is theresult of either -something.But only when structure or arestrictive relative a deicticmarker in the deep Restrictive relatf.ve clause can a KoreanNP be markedwith Id. Chapter Three(3.12). clauses have beenextensively discussed.in i 'this"; ce'that (over The deictic markersare thefollowing: these demonstratives there)1; and ki "that(there)1. The usages of
areillustrated below:
(9) 1 kes-ninpissa-n chayk-iyo
this thing-Texpensive-REL book-is
'This is an expensivebook.'
(10) cechayk-nklEinstein-cenki-iyo 1 that book-TEinstein(ls)-biography-is biography. 'That book overthere is Einstein's
(11) kl chayk-nin eceysasso
that book-Tyesterday bought
'I bought thatbook yesterday.' to what is closeto the speaker, The word i isused with reference the hearer or not(e.g. 9); regardless of whetherit is close to is far fromboth the speaker ce, whenthe referenceis made to what whentheobject.referred tois far and hearer(e.g. 10); _and
In Chapter Two, non-linguistic devices which NP s which aredefinitized ,by - The examples(ch. English are notmarked as definitein Korean. cm:for convenlenc ...41.1
134
(12) a. The roof is leaking.
b. cipung-ka saynta
roof-S leaks
(13) a. The moon is rising.
b. tal-ka ttiko-issta
moon-S is-rising
(14) a. I was driving on thefreeway the other day whensuddenly
the engine began to make afunny noise. I stopped the
car, and when IQpened the hood, I sawthat the radiator
waz boiling.
b. yocennal kosoktolo-111talli nintey, kapcakieyncin-ka
the-other-day freeway-0 ran-and suddenly engine-S
isangha-n so1i-111 nay-kisicakhayessta. cha-111 mec-ko
strange-REL noise-0 make-NOMbegan car 0 stopped- and hus-lil yel-ni latieytha-kakkllhko-issessta
hood-0 opened-whenradiator-S was boiling with the, while All the English nounsunderlined above are marked Notice that the their Korean counterparts arenot marked with subject markers are kainstead of n1n in the mostnatural the translations of the respectiveEnglish sentences, aathough subject NPfs of 9-11--which aredefinitized by a linguisticdevice--
are topicalized inthe most naturaldiscourse.
4.12. .Further.-reStriatiOn.
claimed that definiteness is
condition fOr 135
n 5 & mentionedj. The feature to be not only& definite), but also is not necessarily & mentioned) implies& definite], but the reverse subject NP's usedmaterially aresometimes true.This explains why with ka (not yet marked with nIn(topicalized) and sometimes topicalized). Consider the following:
(15) a. sensayng-ka(na-2A1) ttayliesseyo.
teacher-S me-0 beat .
'The teacher beatme!'
b. sensayng-nin(na-lil) ttayliesseyo
(16) a. cipung-kaecey-puthe sayko-isseyo
roof-S yesterday-fromhas-been-leaking
'The roof has beenleaking sinceyesterday.'
b. cipung-nlnecey-puthe sayko-isseyo said, let us considersome To make clearwhat I have just 15 and 16 environments where onesentence, but notthe other, In home from school,cries for his can be used. When a boy, coming happened at schoolthat parents' sympathy over anevent that had Sentence 15b ispossible only day, he would use15a, but not 15b. already introducedby either when se...nsama'teacher' has been discourse--e.g. afterthe speaker or hearer astopic of the present and he called me boy stated 'theteacher was angryat mv behavior, Suppose an old widow,complaining about to the front andthen -..1 - about herselfpicks out the her relatives'being little concerned She will definitely roof that has beenleaking sinceyesterday.
choose 16a and mightcontinue: 136
(16') a. kIlaytokocheculye-nin salam hana epseyo
but fix-willing-REL man even-oneis-not
'but there is not even oneman whc iswilling to fix it
for me.' already been Sentence 16b is possibleonly when the roof has could mentioned in the presentdiscourse by such astatement as the roof was leaking. In not sleep in thesleeping room, because closely corresponds to fact, ...I. In this respect, TCPZN that only when definitizer marking:I previously claimed definitization is performedby a linguistic deviceis k& 'the' TOPZN--namely, attached to the NP.The same thing canbe said about that en NP can betopicalized only when itis linguistically definitized in the presentdiscourse.6 materially- The next questionto be asked is whether any in used NP must betopicalized if it islinguistically definitized 411,1 another limitation on the present discourse.We seem to ham only an obligatory- current transformationaltheory which employs Some rules can be as optional distinction inrule applicability. as high as low as five percentopticnal while other rules are described as ninety-five percentoptional. If all these are just optionality, we surely are optional, disregardingthe degree of -For native speakers losing some verysignificant generalization. correct but rarelysaid/ know, and will tell us,that one case is 'what is a2most heard; whereas another casemaybe described as
always said/heard'.TOPZN is an optionalrule 137 optionality seems to be verylow--that is, the ruleis almost, but conditioned, but not quite, obligatory.Some non-linguistically here, but the systematic (rule-governed)variation might exist of the investigation of such variationis far beyond the scope present dissertation. (rule 7) indicates that My formulationof the TOPZN rule 7 Basically, this rule has to beordered after PARTICLEINSERTION. A there are two typesof particles: functional and contextual. that shows the functional particle isessentially a case marker 4111111Mv function in the sentenceof the NP to whichit is attached (e.g. eykey 'dative', ka 'subject', 111'object', eyse 'locative', Darticle offers a cueto the ex 'time°,whereas a contextual 'too', contextual relation ofthe NP to which itis attached--e.g. to
man 'only', or ya.'at least'. Both types of particles are Functional particles tranzformationally insertedinto a sentence. those that are insertedby cyclic are dividedinto two subgroups: inserted by pre-cyclic rules (ka and111), and others that are
rules.All these areformulated below: 8 (17) SUBJECT-PARTICLE INSERTION: OBLIG 2 NP, NP V 3> 1 kal 1 2
(18) OBJECT-PARTICLE INSERTION: OBLIG 111, NP, NP, V 7 1 2 1 2 3
(19) DATIVIZATION: 3siNpI ^
138
(20) to-INSERTION:
4 (X, z), CY, Z):=40. 1, 2, 3 + to, 1 2 3 4
The following datasuggest that these variousparticle rules are ordered in the followtng way: (1) precyclic rules that insert functional particles (e.g.19); (2) precyclic rules thatinsert contextual particles(e.g. 20); and lastly (3) cyclicrules that insert functional particles(e.g. 17 and 18):
(21) a. John-man-kahapkyekhayessta
John-alone-S passed (theexam).
'Only John passed theexam.'
b. Mary-ka john-man-JA1salanghanta
Mary-S John-alone-0loves
'Mary loves only John.'
(22) Mary-ka John-eykey-mansenmul-7A1 cuessta.
Mary-S John-tc-alonegift-0 gave
'Mary gave a gift onlyto John.' suggesting that Step 2 above In 21, man precedeska and 111, thus suggesting that Step 1 precedes Step 3; in 22,,eykey precedes man,
precedes Step 2. well as The utilization ofinformation across sentences as
information within a sentencein a-zule formulationenables a 10 grammar to describepresuppositions rather neatlY- Since description of 5 the particle that appearsin the structural 139
particles, and since some -includes both contextualand functional rules (17 and 18), functional particles areinserted by cyclic postcyclic rule. TOPZN must be either alater cyclic rule or a
It will be shownbelow that TOPZN is apostcyclic rule. Kuno (1970: 4.13. TOPICALIZATION as apostcyclic rule. viz. thematic Note I) observed twodifferent uses of Japanese wa, of as, viz. (= topicalized) andcontrastive, and three uses of which seem to be descriptive, exhaustive,and objective--all These are applicable to Koreanwithout radical revision.
illustrated below:
(23) John-nkienehakca-ita
John-T linguist-is
1JOhn is a linguist.* anay-nln enehakca-ka (24) john-nlnenehakca-iciman, linguist-S John-T linguist-is-but he-ofwife-T
ani-ita
not-is
*John is a linguist,but his wife isnot.*
(25) cipung-ka saynta
roof-S leaks
*It is the roofthat is leaking.
(26) cipung-ka saynta
roof-s leaks
'The roof is leaking.* 140
(27) John-nnyenge-ka yuchanghata
John-T English-S fluent-is
zAs for John,his English isfluent.' corresponds to raytopicalized NP Thematic wa (orKorean nin) (or Koreannin)- e.g. 24- I (e.g. 23).As for thecontrastive wa inserted by someprecyclic rulewhich merely conjecturethat it is above conjecturefor the inserts a contextualparticle. I make the contextual informationto Ise it; following reasons: (1) We need to thedelimiting wa (orn_4.n)--which (2) it issemantically closer than tothematic wa (ornin);11 will be discussedin Chapter Five- contrastive wa (orni-n) can occur (3) unlike thematic wa(or 7 only when sentence.12 Thus, ex. is grammatical in anembedded..
John-nn iscontrastiv.:.-Ily used: rnaThayessta hakkyo-ey..ka7nkeskathtako (28) Mary-ninJohn-nLn ecey went-tOlvi.appear-C.F said. MaryT John-DLM.yesterday:,sChool-to have attendedschool yesterday Mary said 'thatJohn appeared to
(although _Tom didnot)t _matter for those To facilitate an,und,_;.,tanding of.the these paralreis readers who donot know Korean orJapanese, I offer from English: Johi (29) a. Johnlikes Mary- b. John. likesMary (= *It,is
*The one' who113,:.-es , : , .,.. . In -.29b,, -Johnis 'uSed-exhausrav
used as an. answer 141
to:a questionlike- 'What did you Ilike29a may beused :as an:answer :In 29a. An exhaustive sayabout jobn2i..John Isuseddescriptively formally-used NP,which carries new qa (orka)-phrase is a amaterially-used NPwhich information; -adescriptiveas-phraseis 22carries theprimary accent, is not yettopicalized.Exhaustive The differencebetween 25 while descriptiveadoes not. is made clearin theirunderlying (exhaustive) and26 (descriptive)
structures251and261below: saynta) ( (k)kes cipungita) (251) ((etten) kes roof-is certain thingleaks the thing
(cipung saynta) chapter. separately discussedlater in this Objective22will be that thematic wain Japanese Kuno (1-12)further observed explain this embedded clause.Heattempted to neveroccilrs in -an-
in thefollowing manner: thematic wa andthe descriptiveas. The diStinctionbetween the clauses.- All as-neutraiizes insubordinate and theexhaustive following"Ethe number three arerealized -asqa,-.shown mine]: is . hihongo-ga dekirukcito o sitte- (30) a. Anata-waJohn ga know *wa JapaneseIs able that You to speak
a.masuka - speak Japanese?* 1Do you knowthat John can John ga suki nako wa Marydesu wa fond isgirl -of ai Jo 142
stating that It cannot bequestioned that Kunois correct in 13 Put thematic wanever occursin the embeddedsentence. have otherwise, TOPZN neverapplies untilallthe cyclic rules to any operated on the finalcycleand thereforecannot apply But it is hard to agree sentences other thanthe topmost sentence. detopicalized in an that a previouslytopicalized NP can become Kuno. embedded sentence, asis implied in theabove quotation from why it should not.However, Of course, thereis no a priori reason unmotivated, but also anunnecessary it, appears to menot only to be the fact is that complication of the grammar.What seems to be topmostone. TOPZN simply maynot apply to anysentence except the But whatabout the other sideof Kuno s observation also neutralizedin exhaustive arid descriptivekat s (or oa/s) are like to suggestthat Kuno may have the embeddedsentence? I should
been misled byinadequate or incorrectexamples. descriptive interpretation for instance,definitely leans toward a ta.i..racof subject.In other words,it is easier to of its embedded ... descriptively used.However, situations where theorphrase is . which the speakermight rare as itis, a situationis conceivable in John notsomebodyelse,that ask his hearer ifhe knew that it was ,John-ga would carrythe accent could speak Japanese,in which case " - . and naturally anexhaustive -meaning. are But the claim thatexhaustive kafs (ormts) e c9as neutral
senteiee .more-.-eas 143
11 for by the situation: where an exhaustiveinterpretation is called
(31) a. anata-waJohn-qa katu-toomoimasu-ka
You-T John-S win-QTthink-QUES
*Do you think thatthe winner will beJohn?*
b. anata-waJcihn-ga okane-o nusunda koto-ositteimasu-ka thing-0 know-QUES You-T John-S money-0 stole
*Do you know thatit was John who stolethe money?*
c watakusi-waJohn-ga katu-to-waomowanakatta
I-T ,john-S thOUghtnot be II,didnot thinkt4iatthe,winner .would John. equivaients:of . _ arguMents can be made forthefcorean. . Exactly :the:spine: . . . _ _ ._ _ the above: (32) a.tangsinnin-:JAnka-.-ikiiila77kla Sayngkakha7seYyo?:.
*Do you think thatthe winner will beJohn?* tangsin-nin John-kakl ton-111humcheka-nkes-111
John-S the money-0 _stole-NOM-0
.aseyyo?,
know-QUES
*Do you know: thatit was John thatstole theMoney?
c. na-ninJohn-ka sayngkakhA-cl, mos- ., I-T John-Swin-Cfr-DLM ,-thirk=NOM 144
renderings, the ka(orm)- As is clearlyshown in the English 4' exhaustively (orformally) phrases above areall subjectswhich are sentence. It used and which arenot neutralizedin the embedded of the usesOf NP in the follows that thereis no neutralization that the non-occurrenceof embedded sentence.It also follows is due not to thematic nin (or wa)in th . embedded sentence postcyclic rule. neutralization, but tothe fact thatTOPZN is a of TOPICALIZATION. 4.14. Restrictions onthe applicability the two problemsof determining In this section,I will discuss obligatory.TOPZN where TOPZN is tobe blocked andwhere it is formal) ka-phrases Why should this is blocked forexhaustive (or answered in thederivations of be so? Thisquestion will be The exhaustiveka and exhaustively, andformally, used NP each other bythe presence descriptive kaaredistinguished from two the beginningof the NP. The or absenceof the mainaccent :on Certainly thedifferent uses of ka arealso: semantically distinct. for this semantic degree of accentcannot beconsidered responsible sentences -by wayof usconsider the following
illustration:
a. sensayng-ka(na-11)ttaYliesseyo
teacher- 'me-0 beat
'The teacher beatme.'t.
b. -sertsaynci-ka-,(na-111) ttaylleiseyO
*It va:s-the--teacher that:- beatme. 145
(34) a. cipung-ka sayyo
roof -S leaks
'The roof leaks.'
b.cfRuni-ka sayyo
lIt is the roof thatis leaking.*
Whereas the (a) sentencesof 33 and 34 aresimple statements, the
underlined NP's in the(b) sentences carry thekeyinformation of The (b) the sentences, as isindicated in the translation. already shown with sentences come fromunderlying conjunctions, as
respect to ex. 5 insect. 4.11.
According to the suggestedanalysissentences 33 and 34 will
have the followingunderlying structures(here irrelevant details 14 areIgnored, such asspeech level ortenses):
(33 ) a. ((sensayng)Np (na) (ttayliessta)v)s
teacher -( ( b. (((etten)salam)Np (na)NP (ttayliessta)v certain man
salam)Npsensayng)(ita) teacher is (saynta)v.)s
-leaks
(((etten) kes)Np (saynta)Y...((ac&)kes) (cipung)Np
certain thing leaks (ita)v)s instance- Of-W-ch. is If PREDICAttPREPOSZEION --PRED PREP) an- 146
NR (Adj-Pr) FORM (ch. 2, rule96)--applies to 33/band 34th, we will have 33b and 34b, respectively; ifRR-FORM (ch. 3, rule 58) applies instead, the respective (c) sentencesbelow will be derived. The respective (b) and (c) sentences in 33 and34 are synonymous with each other.
(33) c. na-lI1 ttayll-n salam-nInsensayng-ita
me-0 beat-REL man-T teacher-is
*The man who beat me is theteacher.*
(34) c. say-nln kes-nin cipung-ita
leaks-REL thing-T roof-is
*The thing that is leakingis the roof.
All that is needed to accountfor ka being marked on a.formally-used
NP in the (b) sentencesof 33 and 34and forn11being used after a relativi_zed NPin the -(c) sentences is to order the two rules
(NR CAdj-Prj FORM andRELATIVIZATION) after PARTICLEINSERTION.15
To be exact, the sharedsubject NPIsin the intro-S1s--the first
conjuncts of331b and341b--are always marked with kaand these
NPIsare not subject toTOPZN because of the=indefiniteness.
the derivation of the(b) sentences of 33 and 34the subject
particle has been originallyinserted by PARTICLE LNSERTION inthe
Intro -S, and later PRED PREP movesthe predicatesof the second
conjuncts into the intro-Vsand replaces the subjectNIDIs of the
original intro-S s with thosepredicates.In the derivation ofthe
(c) sentences, the subjectnps of the second conjunctsof 331b and
341b are 'markedAal 1.47
definitizes those NPIs, to which TOPZN RR-FORM (A) (ch.31 rule 58) suggested above applies. The correctnessof the rule-ordering that PARTICLE INSERTIONis becomes clearer fromthe consideration
a cyclicrule. obligatory, with On the other hand,TOPZN seems to be Consider the following: habitUal and gnomicstatements.
(35) a. inkan-nlnmanmul-ly yengcang-ita
man-T all-the-things-of master-is
'Man is the masterof all the things.
b. olav-rin pholyutongmul-ita
whale-T mammalian-is
'The whale ismammalian. kyohoy-ey kanta c. John-nlnmaywel hanpenssik
John-T every-menth oncechurch-to goes
/John goes tochurch once amonth. marked with ka,then they are If the subject NI:" saboVe can ever be This seems f-o automatically interpretedformally or exhaustively. statements TOPZN iscompletely indicate that inhabitual or gnomic of the the habitual orgnomic character obligatory. In other words, definitization.The statement functionsexactly like linguistic ka-Phrases in theabove sentences automatic formalinterpretation of (viz. can be easilyexplained, since the other#YPe of -Jca to n by obligatory descriptive) willinmaediately be switched
TOPZN. 14 8
The so-called objective ka(or ga)
4.2. By introducing the conceptof a formal NP (an NP moved from the second conjunct into anintro-S) and that of a material NP (an NP used In sentencesother than intro-S), and by restricting TOPZN in such a mannerthat it applies only whenthe
NP is definite, I hope that Ihave neatly described andexplained the'complicated uses of :An vs. ka. But Kuno (1970 Note III). claims that there is a third useofga.(or ka) phrase, which he calls ob'ectiveEa.(or ka). Objective ka is illustrated below:
(36) John-n&n yenge-ka yuchanghata
John-T English-S fluent-is
*John is fluent in English.*
(37) John-nIn toum-ka philyohata
John-T help-Snecessary
*John needs help.
(38) John-nln sul-kamasi-ki silhessta
John-T liquor S drink-NOM disliked.
*John did not feel likedrinking.'
But the term *objectiveka* does not seem to- me tohave adequate
motivation, at least In:Korean,since the verbs yuchanghata
fluent I philyohata
like static adjectives inEnglish 149
361 and 371:
(361) John-ly yenge-kayuchanghatal7
John-of English-S fluent-is
'John's English is fluent.'
(371) toum-ka John-eykeyphilyohata18
help-TJbhn-to necessary
'John needs help.I
TOPZN applies to 361 and371 and derivesthe following:
(36") John-iy-nin yenge-kayuchanghata
(37") John-eykey-nIn toum-kaphilyohata the last one when We need a rule which deletesall particles except
they are In juxtaposition. I will call thisrule PARTICLE-DELETION
(abbr. PART-DEL). PART-DEL is obligatorywhen nIn follows PART-DEL derives 36 and functional particles; ctherwise,optional. difficulties 37 from 36" and 37",respectively. But some serious to derive arise in this first approadh:First there is no way
39a by this approach:
(39) a.na-nLaJohn-kayenge-kayuchanghatankes-113. john-S English-Sfluent-is-NOM-0
mollassta19
did-not-know 150
39a If we follow theanalysis suggested asthe first alternative, post-cyclic rule must be derived from39b by TOPZN.But TOPZN is a this is that there and can be applied only once.The consequence of in 39a.Even if it is no way of switchingJohn-ly in 39b to John-ka there is were possiblefor TOPZN to changeJohn-iy to John-nAn switches nth to ka,in still needed adetopicalization rule which crucial difficulty order to derive 38afrom 38c.A second and more is the derivationof that arises with thefirst approach suggested
40a:
(40) a. na-n&n,john-ka yenge-ka yuchanghata-rAn kes-lll
mollassta fluent in *I did not knowthat it was John who was
English.' John ita))) b. (na mollassta((X.4yyenge yuchanghata)-(x chapter that NP'sused formally I have pointedout earlier in this therefore even if (or exhaustively) cannotbe topicalized; impossible to derive detopicalization werejustifiable, it would be Does this Marl 40a from thesuggested underlyingstructure 40b. analysis and claimthat John in that we have to goback to Kuno's thatyenqe, *English*is 39 and 40 is theunderlying subject and
actually an object,despite the particleka? eems to me The second alternativeanalysis, which
the correct one,suggests thatan,optional rule
changes John-i.y in39b to -jolm-ka in.39a under_ 151
that follow, the(b)sentences seem justifiable?In the sentences a rule whichI to be derivedfrom tha respective(a) sentences by call CASE AGREEMENT;it is an optionalrule.
(41) a. John-nin ki kko1i-111 ccalInta
John- the dog-of tail-10 cuts
'John cuts the tailof the dog.,
b. John-ntn kkay-111 kkoli-lilccalinta
John-T the dog-0 tail-10 cuts
(42) a. ki kay-ixkkoli-ka ccallieciessta
the dog-of tall-Swas-cut
'The tail of thedog was cut.,
b. ki kay4cakkoli-ka ccallieciessta
the dog-S tall-S was-cut
ix sof I in an NP. NP1 CASE AGREEMENTchanges the particle
construction so as to agreewith the particleof the head la? to 111 to agreewith the (= Np.) . Thusin 41 ix is switched changed to ka particle attached tokkoll stall' . And in 42 ix is to the head nounkkoli.Thus, to agree withthe particle attached with the respective the (a) sentencesof 41. and 42 are synonymous
(b) sentences.Compare these withthefollowin
(43) a. John-ran k&kay-ix pap-1.&1epcillessta turned-over (44) a. k& pap-ka epcilleciessta
the dog-of food-S isturned-over
*The food of the dogis turned over.* b. k& kay-ka pap-kaepcilleciessta
the dog-S food-Sis-turned over derived from therespective The (b) sentencesof 43 and 44, which are
(a) sentences by CASEAGREEMENT, are notgrammatical The only and 44 on the difference between 41and 42 on the onehand and 43
is that NP. and NP. arein an inalienablerelationship other hand 3 20 and NP. must be only in 41and 42. In other words NP. CASE AGREEMENTto inalienable relationwith each other inorder. for *dog* andREE*food' apply to NP. NPconstructiOn. Thus,kay explains the are not In aninalienable relation, which
ungrammaticality of 43band 44b. Exs.43b and 44b aregrammatical, that however, if the NP.(= kay in this case)isformally used; 3_ formally used INEDi any CASE AGREEMENT seemsto be applicableto a between NP. and NP.is time regardless ofwhether the relation 3-
inalienable or not. "that arise with In this secondapproach the difficulties to arise.The first the suggested firstapproach naturally cease to whichCASE difficulty is the derivation.of 39h from 39a, syntax--supplies asimple AGREEMENT--independently neededin Korean of derivation. The seconddifficulty has to dowith the derivation ready solution here: 40a from 40b.CASEAGREEMENT supplies another derive 40c from 40b;end CASE PRIM PREP andCOMPLEMENTATION will Ita, 153
CASE AGREENENTis applicable AGREEMENT derives40a from 40c, since construction.
yuChanghata-n1n kes-111 c. na n1nJohn-y yenge-ka
I-T John-of English-Sfluent-is-NOM-0
mollassta
did-not-know 'objective ka 1discussed The third caseof the so-called below. Consider thefollowing: silhessta (381)` a. John-nkisul-l&l masi-ki(ka)
John-T liquor-0 drink-NOM-Sunlikable
'John did not feellike drinking.
b. John-nlnsul-ka mas1-4c1(ka)siihessta unlikable - John=T liquor-S drink-NOM-S silhessta c. *John-n1nsul 111 (or-ka) masi-ki-111 action-verbals take oand state-veebals Kuno (III-9)claimed 'that and went on tocharacterize theveeb mi-ta 21 forobject marking' He seems toimply--correctly, I 'want to seeas astate-verbal. verbal governsthe whole think--that thefeature of the last combined andfunction like complex verb .whentwo or more verbs are of verbals whichfunction as a a simpleverb. This combination verb. Furthermore,I will simple verb I willcall an a9glutinated combines withother verbs call agglutinativeverbal a verbal which verbals are and behaves like asimple verb. All agglutinative verbals intransitive.21 Another characteristicOf agglutinative a -veNP*S ar 154
the person correferential--in otherwordsthe personwho drinks and correferential with eachother. in 38*b. who hates todrink are the nominalizerki The verb silhta*bedisliked' can have dative NP s--cf.381e--are only when embeddedsubject and matrix the ungrammaticalityof correferential witheach otheras shown by
n_4.n kes that* even when 45b. But silhta canhave the norninalizer by the- grammaticalityof 45a they are notcorreferentialas shown below: kes-ka silhessta (45) a. John-nlnMary-ka sul-lll disliked John-T Mary-S liquor-0 drink-NOM-S
*Johndid,not likeMary's_ drinking. masi-ki-ka silhessta b. *John-ninMary-ka sul-lll S disliked John-T Mary-Sliquor-0 drink-NOM
**John did notlike John's drinking. agglutinative verbalsuch a Since silhtawith nIn kes is not an will never InEluenceth feature asr; transitive] of siihta and therefore46b preceding verbwhen nn kesis its nominalizer;
is impossible: silhessta (46) a. John-nln masi-nki kes-ka
John-T.Liquor-0 drink-NOM-Swas-disliked
*Johndisliked drinking.* silhessta *John-n.4.n. sul-kamasi-nInkes-ka
of 38 with aview to putting I will discussthe derivation 155
381b is 381d below: underlying structureof 381 a and
(381) d.
structures from thisunderlying In order toderive the surface
DATIVIZATION(rule. 19)
PARTICLE INSERTION(rule 17 and 18)
TOPICALIZATION(rule 7)-
PARTICLE DELETION(cf. 4.2)
REFLEXIVIZATION'(appendix I rule 7)
COMPLEMENTATION
PREDICATE PROMOTION PROMOTION areindependently needed All these rulesexcept PREDICATE rules have alreadybeen diScussed. in Korean syntax.The first four matrix NP,REFLEXIVIZATIONeither If the NP iscorreferential to a -self/ or completelydeletes it changes an embeddedNP to caki 2 does not resultthereby.2 under the conditionthat ambiguity
COMPLEMENTATION rule PREDICATE PROMOTIONis an optional minor
frequently inagglutinative languages r. 156
predicate and amalgamatesit PROMOTION simplyraises an embedded complex verb whichbehaves like with the matrixpredicate to create a the complex verb. one verb.The features ofthe second verb govern
This is formulated as4. (abbr. PRED PROM): (47) PREDICATE PROMOTION
xrr, 13 1 2 is a precyclicrule and I have previouslyshown that DATIVIZATION postcyclic rules.The rest that TOPZN andPARTICLE DELETION are applicable in thefollowing order: of the rules arecyclic and are COMPLEMENTATIaN FREDPROM. PARTICLE INSERTION,REFLEXIVIZATION,
These will beillustrated with 38. and derive 38te: First, DATIV1ZATIONwill apply to 38Id
(38') e. S NP NP V.
S John eykey silhta s.,--" NP NP "V I I Jdhn sul masita INSERTION isapplicable and In the firstcycle only PARTICLE
derives 381f from38te:
(381) f. 157
and On the secondcycle, PARTICLEINSERTION REFLEXIVIZATION,
COMPLEMENTATION applyand derive 381gfrom 38tf:
(389 g. NP V
John eykey silhta
NP V ka
sul 3.11 masi ki not apply, if it Since PRED PROM is anoptional rule, it may or may
applies, 38th will bederived after S/ hasbeenrmuned:23
(389 h.
John eykey
masiki silhta
will apply and derive38ta Postcyclic TOPZN andPARTICLE DELETION
and 381b from381g and 381h. are Thus, all three casesof so-called'objective kat Now .we do not haveto shown to beunderlying sub ject markers that ka always is asubject weaken or compromisethe generalization between marker and lil is anobject marker.The difference shown to correspondto the exhaustive ka anddescriptive ka has been The analysissuggested formal vs. materialdistinction in the NP. distinction between the use in this chapter alsoindicates that the to ka- of exhaustive ka andthat of descriptiveka is not restricted
phrases This isshown below: 158
( 48) a. na-n1n Mary-111salanghanta
-I-T Mary-0 love
/I love Mary/
(m. Mary salanghanta)
b. na-nAn Mary-111salanghanta
I-T Mary-0 love
/It is Mary that Ilove./
(na X salanghanta)-(XMary ita)
I X love X Ma.ryis
(49) a. John-lytongsayng-ka iiciessta
John-of brother-S has-won
/John's brother haswon./
(John-ly tongsayngikiessta)
b. John-ly tongsayng-kaikiessta
/It is3Shnts brother that has won./
(X-ly tongsayngikiessta).(X John ita) rAn has been shownto As far asn1n-phrases areconcerned, thematic contrastive n1n hasbeen claimed tobe a be derived byTOPM, and precyclic rule. contextual particlewhich derived by some 159
CHAPTERFIVE:NEGATION
Introduction
to shed 5.1. The present analysisof Korean syntax seems great deal some light onanother problem whichhas recently drawn a In of attention from nativegenerative syntacticians--Negation.
Korean, there are two waysof negating sentences,illustrated below:
(3.) a. ai-ka canta
child-S sleeps.
'The child sleeps.'
b. ai-ka ani-canta
child-S not-sleeps
wrhe child does notsleep.'
c. ai-kaca-c1(1&1) ani-hanta
child -S sleep-NOM-0not-does
not sleep. . 'The child does
(2) a. Mary-n&n yeypp-ita Mary-T pretty *Nary ispretty.' b Plary-nnani-yeypplta Mary-T not-pretty 'Mary is notpretty.'
c. Mary-ninyeypp-l-ci ani-hata
Mary-Tpretty-NOM not-does 160
the verb, whereas In the (b) sentences, NEGis simply posited before in the (c) sentences threethings happen: (1) the original verb is nominalized--always with ki 'to'. (2) The verb hata Idol isused as a verb of the mainclause.This verb--viz. hata--whichis used in negating sentences is adifferent lexical item from atransitive verb hata Idol.The difference between thesetwo verbs will be discussed in 5.32. (3) NEG is added before hata. I will call the negation which is examplifiedby the (b) sentencesabove Type I negation and the sort thatis examplified by the(c) sentences,
Type II negation.
Basically, three approachesto negation in Korean are conceivable: (1) an analysis whichpostulates separate underlying
structures for both types;(2) an analysis which adopts adopts NEG- NEG-TRANSPORTATION (abbr.NT); (3) an analysis which
INCORPORATION (abbr. NI).The NT approach and theNI approach
will be extensivelydiscussed in 5 2. and 5.3respectively. different The separate-underlying-structureanalysis postulates two According underlying structures forType I and Type IInegations.
to this analysis, ltband 11c belcw would be theunderlying
structures of lb and lcrespectively:
(11) b. 161
NP VP _____--T-..-...... _.... NP Neg V
1 1 1 S ani hanta
NP VP
ai V
ca(ta)
Song (1967:58-60) hassuggested such analyses. But in view of the fact that the two types arecompletely synonymous witheach other it appears that thereis no motivation forpostulating separate deep structures.
The NEG-TRANSPORTATIONapproach
5.2. The NT approach has beensuggested by Lee (1970:175 et seq.).Lee (177) postulates 3 asthe underlying structure for both lb (TypeI) and lc (Type II), justifyinghis position by claiming that 'at present,there is no well-motivatedrule' by which sentences like lb canbe derived from 4:
(3)
NP 162
(4)
ha
NP VP
al ca discussed in 5.22. This analysis of thederivation of lb will be choice but to raisethe For the derivationof lc from 3, Lee has no where NEG NEG into thehigher sentence fromthe embedded sentence, he uses NT.But the is originallypostulated. For this purpose illegitimate. use of NTwill be shown belowto be absolutely In the first 5.21. The illegitimacyof NT with hata. An example of a place, hata is notwhat can be called aNT verb. There are two readingsof NT verb issayngkakhata Ito think'. These two sayngkak-hata:a NT readingand a non-NT reading. different readings are syntacticallymarked in Korean with
nominali7ers: ko with aNT reading and n_4.nkes with a non-NT illustrate the tworeadings: reading.The following sentences
(5) a. na-nin (John-katases sikanpakkey(n.ln) kongpuhayessta-ko)
John-Sfive hours other-than-DLKstudied-QT
sayngkakha-ci ani-hayessta
think-NOM not-did than five hours.' did not think thatJohn studied more 163
(5) b. na-nin (johnAcatases sikanpakkeytnin] kongpu-hayezsta- studied- I-T John-3five hours other-than-DLM
nAn kes-lil) sayngkakha-ciani-hayessta
NOM-0 think-NOM not-did
c. na-nln(John-ka tases sikan mankongpu-hayessta-nin kes-
I-T john-Sfive hours only studied-NOM-
lil sayngkakha-ciani-hayessta
0 think-NOM not-did five hours. II did not rememberthat John studied only reading--viz. 5a--the speipker With sayn9kakhataIthinkt in its NT value of a certain is giving his opinion orjudgment about the truth statement, while with theverb in its non4FTreading one claims Naturally, one's involvement in themental process ofthinking. sayngkak-hata in itsNT-reading cannot benegated in the deep structure. For if one makes nojudgment--i.e. if the verbis that the negated--no object ofjudgment is needed: I suggest, then, 2 ynderlying structure of5a is 5ta not 51b:
(51) a.
NP
NEG
John-ka tases sikanpakkey(nin) kongpu hayessta 164
(5 ) b. V NP 1 NEG
NP r 1 na sayngkakhata
kongpu ayesta John-ka tasessikan pakkey 5Ia insteadof 51b as the The correctnessof postulating that clearly demonstratedby the fact underlying structureof 5a is only in a use pakkeytother than' In Korean itis possible to 6b, which haspakkey in au negative sentence:Thus, sentence affirmative sentence,is ungrammatical. pakkeykongpu-ha-ciani-havessta (6) a. john-nintases sikan other-than study-NaAnot-did john-T five hour
'John studiedonly fivehours.' kongpu-hayessta b. John-nlntases sikanpakkey
'John studiedonly five hours. sikan mankongpu-hayessta c.John-nln tases
'John studiedonly five hours. sentence of6a--which isaffirmativehas Notice thatthe embedded NEG-TRANSPORTATIONmust pakkey": In order to derive5a from 51a, This rule isformulated asfollows: therefore beapplied. (abbr. NT) (7) NEG-TRANSPORTATION
ii[EYISINP2 NEG1S]NPI 2- 165
grammaticality of 5c That NT is an optionalrule is shown by the below--where NT has not appliedto 5ta--and by the synonymyof 5a with 5c:
(5) c. na-nin (john-ka tasessikanpakkey kongpuha-ci and.-
I-T John-Sfive hour other-thanstudy-NOM not
hayessta-ko) sayngkakhanta
did-QT think
'I think that Johnhas studied only five hours.' from 3, it To substantiate Lee'sclaim that NT generates lc is necessary to show thathata is a NT verb. But the following facts clearly indicate thatit is not a NT verb: (1) If hata is a
NT verb, and if 3 isthe correct underlytngstructure of lc, then
ld below--where NT hasnot applied to 3--shouldbe grammatical, 3 since NT is an optionalrule.But ld is not grammatical:
(1) d. ani ca-ki(111) hanta
child-S not sleep-NOM-0does
'The child does not sleep.' verbs are not required (2) The followingfeatures of Kartunnen's A
for Korean NT verbs. Kartunnen (1969) dividesverbs into two 4 in the predicate of a groups: A and B. When an A verb occurs obtain.: (a) The subject matrix sentence, thefollowing conditions sentence are necessarily of a matrixyerb andthat of a constituent
the same; (b) the embeddedverb is tensele sin the deep structure; embedded sentence; (c) time adverbialscannot remain in the verb implies the samealso (d) negating orinterrogating the matrix
175,, 166 with respect to the embeddedverb, etc.These characteristics are illustrated in comparison of 8(an A-verb) and 9 (a B verb):
(8) a. *John managed (for) Mary to run.
b. *John managed to have run.
c. *John managed to seeMary tomorrow.
d. *John did not manage to seeMary, but saw her.
(9) a. John expected Mary to run.
b. John expected Mary to have run.
c. john expectedto.see Mary tomorrow.
d. John did not expect to seeMary, but saw her.
Compare these with thefollowing Korean sentences, allof'which have
hata as their matrix verb:
mek-ki-to) hayessta (10) a.*John-Mn (Mary-ka s
John=T Mary-S eat-NOM-toodid
* *John did Mary's alsoeating.
b. *John-nln ka-ass-ciani-hata
John-T go-PAST-NOM not-does
*,John does not went.'
c. *JOhn-ninneyil Mary-JAl manna-ki-tohayessta
John-T tomorrow Mary-0 meet-NOM-toodoes
* 1John did meeting Marytomorrow.1
d. *John-n&n (Mary-1A1manna-ci) ani-hayess-&na
John-T' Mary-0 meet-NOM not-did-but
Mary-II1 mannassta
Mary-0 met' 4- * tjohn did not meet Mary,but met Mary.' exr, 167
Now, the crucial The verb hatabehaves exactly like Averbs. to the class of Averbs question here iswhether NT verbs belong compare thebehavior of or to thatof B verbs. We may now the characteristic sayngkak-hata 'thinks--atypical NT verb--with
behavior of .A verbs. sayngkakhayessta (11) a. John-nln(Mary-ka mek.41ila-ko)
John-T Mary-S eat-QT thought
'John thought Marywould eat.I sayngkakhanta b. John-nIn(Plary-ka ka-ass-Ililako)s
John-T Mary-S went-QT thinks
'John thinks that Marywent.' manna-lilako)s sayngkakhayessta c. John-n&n(neyil Mary-111 thought John-T tomorrow Mary-0meet-QT
'John thought thathe would meetMary tomorrow.' sayngkakha-ci ani- d. John-nIn(Mary-lil manna-lilako)s not- John-T Mary-0, meet-QT think-NOM
hayessIna, Mary-111mannassta
but Mary-0 met to meet Mary, but 'John did not thinkthat he was going
he did meet her.' This fact showsthat the NT Ex. 11 exactlyparallels 9 (not8). Therefore hata--whichis an verbs belong tothe clasS of B verbs.
A verb--is nota NT verb. In utilization ofthe NT rule Another argumentagainst'the by the negation (whichis illustrated the derivationof a Type II
4'77' 168
-derivation of lc from 3) is related to the analysis of a double
negation. Consider the following:
(12) John-nin ani mek-ci-lil ani-hayessta
John-T not eat-NOM-0not-did
'It is not the case that he did not eat.'
Lee (194-6) suggests that 12 has 12' as its underlying structure:
(129
NP VP
John NP Neg V i I I S ani ha
NP VP ./"- John Neg V
mek
If NEG is allowed to be attached to the matrix verb hata in 12',
wlw not also in lc?The reason offered by. Lee (177) for such a
distinction--viz, there is no well-motivated rule to derive a
sentence like lc from a structure like 4--does not seem sound.
For NEG-LOWERING, which lowers NEG from the matrix sentence into
the embedded sentence (e.g. the deri7ation of lc from 4), is as
well motivated as NEG-RAISING, which is suggested by. Lee's NT
analysis. Besides, it is now accepted:by most analysts thatNEG
comes from a higher predicate in the deep structure.This lends
an additional support to NEG-LOWERING. Independent justification
for the higher-predicate analysis of NEG will be given later.
Till now, I have shown that the NT approach is not the
correct analysis of negation in Korean by showing that hata is not 169
. aNT verb:With hata used inTYpe II negation, NT should be
obligatory, although NT isactually optional with otherNT verbs;
and hata behaves like Averbs, although other NTverbs belong to
the class of B verbs.Another argument againstthe NT approach is
related to the derivationof lb from 3. In order 5.22. The theoreticalimplications of ha-DELETION. the to derive Type Inegation, one needs a ruleto delete hata, since deep structure of NT approach postulatesthat hata is present in the This rule of ha- every singlesentence--as illustrated by3. Lee (185) DELETION is responsiblefor the derivation of lbfrom 3.
formulates ha-DELETION inthe following manner: 5 (13) ha-DEL-2 OBLIG X -[X-V] - ha - X >1 2 3 g 5 S s
hata in the deep Before I discuss theimplications contingent on
structure of everysentence and the applicationof 13 to derive formulation of rule 13. Type I negation, Iwill comment on the
Immediately after presentingthe rule, Lee adds that'the feature
specificationrviz. j.. Nnis necessary in ordernot to delete the
verbs which verb hata of a denominalverb'.Korean has a group of to Japanese suru--to are formed byadding hata--which corresponds
some nouns.Thus, kongpu /research'is a noun, but konTou-hata
'to study' is a verb.For another example,tacenq 'kindness' is Lee calls these a noun, buttaceng-hata'ito be kind' is averb. formulation; Lee verbs denominal verbs.In the above rule hata; the latter is identifies verbalizinghata with negativizing 170
'lazed as the higher predicate of a Type II negation. However, the
feature r+N]ris rather peculiar if viewed as an inherent property L- of a verb--it is hard to allow aE.,N] verb in our theory.
Furthermore, the difference betweenthe two hata's is revealed inthe
negations of sentences whichcontain them.Consider the following:
(14) a. John-nin uphyo ani-hanta
John-T stamp collect-NOM-0not-does
'John does not do stamp-collecting.'
b. John-nin uphyo mol-ci-lilani-hanta
'John does not do stamp-collecting.'
(15) a. *John-nin uphyo 111moi-ki(111) ani-hanta
John-T stamp-0 collect-NOM-0 not does
'John does not collect stamps.
b. John-nnuphyo-lli mo-cl-111 ani-hanta
'John does not collect stamps.'
That hata in 14 is averbalizer is shown by its comparisonwith hata
in 16:
sucip-111 ani-hanta . (16) a. john-nn uphyo
john-T stamp collection-0 not-does
'John does not do stampcollecting.'
b. john-nlnuphyo-lilsucip-hanta
John-T stamp-0 collects
'John collects stamps.'
Sentence 16a is semanticallyidentical with 14a.The only difference
between 14a and 16a isthat 16a has the Sino-Koreanmorpheme
1 CO 171
.:;ucin 'collection' in place ofthe Korean nominulizedform m-J'i
'collecting', which occursin 14a. All Sino-Korean wordsfunction to be as nouns, inorder to function asverbs, verbalizing hata has
suffixed to them.And In Korean there is arestriction that no
denominal verb is derivable from a pureKorean nominalizedform.
Therefore, 14a cannot betransformed into 14c below:
(14) c. *John-nth upyo-111mo&-ki ani-hanta
John-T stamp-0collect-NOM not does
'John doesnot collect stamps.' function like a The compound word mol-kihata 'collect' does not
single verb, as sucip-hata'collect' does. Anyhow, hata in both verbalizing hata which 14a and 16a is clearlythe same rorpheme as
makes kongpu 'research'into a verb.The difference between 14 and negativizing hata--e.g.15--and verbalizing hata--e.g. hata does 16--is then the following:Only before negativtzing This is the nominalizer kichange to ci by amorphophonemic rule. changed shown by theungrammaticality of 15a.But ki cannot be ungrammaticality of 14b. before verbalizing hatal asshown by the
This idiosyncracy ofnegativizing hata is further
illustrated by the followingexample:
(17) a. na-nth John-ka iltthgha-ki-nth ani-palanta
I-T John-Stop do-NOM-DLM not-hope
'I do not hope thatJohn will take thefirst place.'
b. *na-nth John-kailtIng ha-el-nth ani-palanta
'I do not hope thatJohn will take thefirst place.' 172
(18) a. John-ka ilting ha-ci-ninani-hayessta
John-Stop do-NOM-DLM not-did
*John did not take the firstplace.*
b. JOhn-ka ani-hayessta
*John did not take the firstplace.*
The idiosyncracy of negativizinghata that ki changes to ci in negative sentence when it is usedis ignored in Lee's formulation of rule 13.
What then is the theoreticalimplication of postulating hata as a higher predicate of everysentence?Actually, a sentence with triple negation isgrammatical, e.g. 19:
(19) Speaker A: John-ka 1 yak-111 mek-ciani-ha-myen
John-Sthis medicine-0 take-NOMnot-does-if
etteckey ha-lkkayo?
what do-shall
*What if John will not take thismedicine?*
ani-ha-ci mos-ha-ci ani-hlkkayo? Speaker B: John-ka mek-ci OMEN=
John-Seat-NOM not-does-NOM not-does-NOMmt.. do-shall
*Wouldn't it be unreasonable to expectthat John
would not take this medicine?*
If we pursue Lee'sreasoning--that since hata appearsin negation,
it has to be in the deepstructure of every sentence,whether
negated or not--then becauseof 1932 every sentencehas to have at
least three levels of embeddingwith three higher hata verbsin the
deep structure. In other words, since19B would have 19v3 assits 173
its underlyingstructure: underlying structure,20 must have 20' as
(19') B.
NP 212-......
I S Neg V I I NP VP ani ha(lkkayo)
Neg Ir
VP mos ha(ci) ,40/\ Neg V
NP NP VP ani ha(ci)
JOhnI yak°\ V 1 mek(c1)
(20) John-ka 1 yak-lilmek-Ilkkayo?
John-S thismedicine-0 eat-will
Will John take thisdrug?*
(201)
NP VP V
NP VP ha(lkk)
NP VP ha N, NP NP VP ha 1 John V mek 174
This would mean that everysingle sentence had to have at least
three hata verbs in its deepstructur,11 Three is of course an
arbitrary limit. Theoretically, there is no limit tothe possible
depth of negative embeddings,and so a sentence may have an
unlimited number of hatals inits deep structure. The NT approach
(which suggests postulating hata asthe underlying higherpredicate
of every sentence and deletinghata in affirmative sentences)is
simply incorrect.To reiterate the points made sofar, the NT
approach cannot be correct becausenegativizing hata is not an
NT verb and hata-OELETION causesthe aWkward problems discussed
above.
TheNEG-INCORPORATION approach
5.3. The lal:t possible way ofanalyzing negation in Korean--
which I will show to be thecorrect one--has NEG as the nexthigher
predicate. Both lb and lc (which arerepeated below for convenience)
will have 21 as theirunderlying structure in thisanalysis:
(1) b. ai-ka ani-canta
child-S not-sleeps
'The child does not sleep.'
c. ai-kaca-c1(1W ani-hanta
child-S sleep-NOM-0 not does
'The child does not sleep.'
(21) (((ai canta) NEG) PRESENT)
child sleep
Because of the irrelevancy oftense here and in the interestsof 114 175
.simplifying the exposition, I have so far ignored the status of
TENSE in the discussion of underlying structures.But I will
discuss TENSE from now on, since it is directly relevant to
negation in Korean. I will still ignore the performative marker
(particles distinguishing declarative, interrogative, and
imperative sentences), which is ultimately derived from a
performative verb (for details, see Lee 1970, ch. 2). The tree
diagram for 21 is given below:
(219 V IP
1 PRESENT
ani
1 ai ca
In order to derive surface structuresfrom 21, we need the
following two rules:
(22) NEG-INCORPORATION (abbr. NI) OPT [X, V3 NEG 1, 3 + S2 1 2 3
(23) ha-ADDITION ( abbr. ha.-ADD1T)6
[1 Y.NEG CONTEXTUAL PARTMLET Jv 2
OBL1G 2 + ha
NI is a rule similar to ADVERBIALIZATION(abbr. ADVB), which lowers
a higher predicate into alower clause, making it also an adverb.
ADVB is illustrated in the derivation. of 24a from24hhere:7 176
(24) a. Jonn-ka pplalikey tallinta
john-S fast runs
*John runs fast.*
VP V
John ADV V
ppalkey tallinta
ppalita) PRESENT) b. (((John tallinta)s s John run fast
NP
PRESENT / NP AT
ppall
NP
John talli
The ha-ADDIT rule is neededwhen NEG is to be realized as the surface verb--1.e. when NEG has notbeen lowered by NI- or when contextual particles (e.g. to*too*, ya *at least*, manonly*) are
attached to the verb. The latter is illustrated bythe derivation
of 25a from its underlyingstructure25b:
(25) a. John-nln. Mary-1A1 mannassta. (kiliko) John-nln Marv-2A1
John-T hary-0 met John-T NarY-0
chacaka-ki-to hayessta
visit-NOM-too did
*John met Mary, and he evenvisited her.* 177
411 (25) b. ((John Mary mannata)sPAST).((john Mary
meet
chacakat a ), PAST)
visit PAST).((John Mary c. ((John Mary mannata)s
chacakota-to) PAST)
too
INSERTION (ch. 4, rule20) One instance ofCONTEXTUAL PARTICLE particle to ttoot to applies to 25b andattaches the contextual satisfies the verb. The resultant derivedconstituent structure description of 23. Therefore 23 the second part ofthe structural from 25c after applies and attacheshata.Then 23 generates 25a 8 SPELLING. the application ofCOMPLEYXNTATION and TENSE If NI applies 5.31. The correct derivationof negation. will be derived: to the underlyingstructure 211 ex. 26 (26) NPV
1 PRESENT
NP
ai NEG V
ani cata
(Ch. 41 rule 17) will Then TENSE SPRIN,INGand DARTICLE INSERTION not apply to 21, finally derive ibfrom 26. If optional NI does generate 27: ha-ADDIT willautomatically app7y and
187 178
( 2 7)
NP V
1 1 PRESENT
NP V
NEG V
.71%N 1 NP V ani ha
1 1 al ca
COMPLEMENTATION (cf. 4.2) and TENSESPELLING will generate lc from
27.
5.32. Justification of the NEG-INCORPORATIONanalysis.
First of all, the infinite-regressproblem--which was discussed in
5.22 as a difficulty of the NTapproach--does not arise with NI analysis. Thus, sentences 19B and 20 willhave 19"B and 20" as their underlying structuresin the NI analysis:
(19") B.
NP V 1 TENSE (ilkkayo)
1 1 NEG
NP V ant
1 1 NEG
V moS
NEG
NP LT V ari. f John i yai mek
'GO 179
(20")
NP V
TENSE (ilkkayo) NP NP V //\ 1 John1 yak mek than 20", just as28a has more The diagram 19"Bhas more embeddings embeddings than 28b:
(28) a. John knewthat he would passthe exam.
b. John will passthe exam. and generate 29,from which The ha-ADDIT rulewill apply to 19"B
COMPLEMENTATION and TENSESPELLING will generate193:
(29)
NP V
1 1 PRESENT (ilkkayo)
NEG ha
1
NEG ha
NP V mos
1 NEG ha
1 NP NP V ani
John yak mek negativizing hata isexplained Second, theidiosyncracy of followed by a In a natural manner.Only before a NEG ddes ki become ci. In other words, transformationallyinserted hata 180
exicon- and hata hata which is in the l 7 there are two hatals: 1, 2' which is not in thelexicon, but added to asentence by a
9 Only before hatadoes ki become ci. The transformation. 2 derivations difference between hataand hatais illustrated in the 1 2 of 14a and 15b, which arerepeated here forconvenience:
(14) a. John-nln uphyomc4-ki-11 ani-hanta
John-T stamp collect-NOM-0 not-does
'John does not do stampcollecting.'
(15) b. John-nin uphyo-111mc4-ci-111 ani-hanta
John-T stamp-0 collect-NOM-0 not-does
'Jam does not collectstamps.'
sentences are given below: The underlyingstructures of the above
(140 NP
PRESENT
NP i i S2. NEG ..,°% /. hataani NP NP 1
I I John. S, oui:YrLoki. 181
(15I)
NP V 10 PRESENT
NP V
NEG
NP NP V ani
John uphyo molta
If NI applies to 14', then 14a will ultimately be generated; if not, 14c will be generated:
(14) c. John-nIn uphyo mol-ki-111 ani-hanta
John-T stamp collect-NOM-0 doTNOM not-does
'John does not do stamp collecting.'
If NI does not apply to 15', then 15b will be derived; but if it does, 15c will be derived:
(15) c. john-nln uphyo-7A1 ani-mointa
John-T stamp-0 not-collect
'John does not collect stamps.'
There is no way to derive 14d and 15d. They are therefore ungrammatical:
(14) d. John-nin uphyo ani mo&-ki-111 hanta
'John does not do stamp collectinc.'
(15) d. John-nln mo-ki-111 ha-ci ani-hanta
1John-T does not collect stamps.'
Third, the NI approach enables us to keep the generalization that the lexical item hata doill,(1421 is a verl. which does not rsl 182
have a sentential object. It might seem that 14, constitutes a counterexample to the foregoing, but in fact S3 in 141 must be an idiom, which functions like a single lexical item rather than a sentence. Therefore, a sentence like 14, is not a counterexampae to the generalization just given.Consider the following:
(30) a. uli-nin macchu-ki-111 hanta
we-T quiz fill-NOM-0 do
11We play a quiz game.,
b. ull-nin quiz macchu-k1-111 ani hanta
(31) a. John-nln 1500-mi ani hayessta
JOhn-T 1500-meter run-NOMO not did
*John did not participate inthe 1500-meter race.,
b. oJohn-nn olay ani hayessta
John-T long run-NOM-0 not did
*John did not run long.,
The nominalized phrases can be objects of hatal only when In idioms or when used as names of games or hobbies, as in the underlined words of 30a, 30b, and 31a. As the nominalized phrase of 31b is not used as an idiom or as the name of a game, hata in 31b cannot be hate,. If hate in 31b were hata2, then nominalizing ki should have been changed to cl. Furthermore, such naming does not in Korean allow particles within or between Sino-Korean words nominalized forms used as idioms, or their combinations. The (b) phrases below are all odd: 192 183
(32) a. cencayng yukacoktouw-ki cukan
war the bereft help-NOMweek
'The week for helpingthe war-bereft.'
b. *cencayng-lyyukacok-111 touw-ki-ly cukan
war-of the bereft-0help410M-of week
(33) a. tayhanminkuktoklip kinyem 11
Korea independence celebrationday
'The Independence dayof Korea.'
b. *tayhanminkuk-lytoklip-111 kinyem-ly ii day Korea-of independence-0 celebration-of phrase, a forced If 31b had lacked111 in the underlined interpretation--viz. that Johndid not play a gamecalled But as it is, olay-talliki 'longrace'--might have beenpossible.
31b is completelyungrammatical. commented on the Lastly, Song(1967:59-61) has correctly awkwardness of 34 and 35below, in contrast withthe grammaticality precise nature of the of 36 and 37, althoughhe did not 'know the hearrd] (1) and (ii) discomfort felt by nativespeakers when they and readily in 2rmy 34 and 353, which areperfectly grammatical to, ya., acceptable...with otherparticles like n1n,
(34) *pi-ka o-ki-111hanta
rain-S come-NOM-0 does
'It rains.'
(35) *nalssi-ka ch-ki-ka hatef
weather-S cold-NOM-S does
'The weather is cold.' 184
(36) uli-nin pi-ka o-ki-111kitalinta
we-T rain-S come-NOM-0 wait 0 tWe wait for rain tocome.'
(37) nalssi-ka chuw-ki-tohata
weather-S cold-NOM-too does
'The weather is extremelycold.t neatly accounted for The ungrammaticalityof 34 and 35 has been when contextual by my formulation ofha-ADDIT.For hata is added particles are attached to averb, or when NEG isthe only constituent of a verb (cf. rule23). Structures like 34 and35 from the do not satisfy the structuraldescription of ha-ADDIT outset, and so hata in 34 and35 cannot be "..lata2,which is hatall which is transformationally inserted. Can the hata there be
in the lexicon? Subjects should be agentswith hatal, but
train' (subject in 34)and nalssi 'weather'(subject in 35) be hatal, are not agents,and therefore hatain 34 and 35 cannot be motivated at either. In other words, hatain 34 and 35 cannot
all. That is why 34 and35 are not grammatical.
194 185
APPENDIX I:FORMAL CONSTRAINTS ON THEDELETION
OF REFLEXIVE PRONOUNS
G. Lakoff (1971:340) arguesthat 'the principlesgoverning the distribution of morphemeswill involve presuppositional phenomenon: information'. He illustrates hisclaim by the following
(1) a. I get my paychecktomorrow.
b. ?*I get a cold tomorrow.
(2) a. The astronauts returnto the earth tomorrow.
b. ?*The astronautsreturn safely tomorrow.
(3) a. Sam gets a dayoff tomorrow.
b. ?*Sam enjoys his dayoff tomorrow. sentences are in the Although all the eventsdescribed by the above
future tense, for some reasonthe deletion of willis allowed only explanation seems to be with the (a) sentences.The only possible will can be deleted. that only when a futureevent has been arranged, describe events In other words, the(b) sentences of 1-3 normally deletion of will which are not arrangedbeforehand: .therefore the unusual situation such aswhen it is blocked. However, in a rather cold germs in a has been arranged for meto be injected with The conclusion medical experiment, lb canbe perfectly grammatical.
is that grammar cannotbe presupposition-freewithout renderino
many importantgeneralizations imporsible. for Another case.wherepresuppositAri is absolutely.needed
be mentioned. In Korean, as inJapanese, syntactic: explanation'may . 195 186 pronominalization by deletion is oneof the major processes which have made the so-called'syntactic component' unmanageably complicated, simply because thetheory has not allowed presupposition in syntax. Consider the following:
(4) a. ?*John-nin caki-kaShakespeare-lll caki-iy chinku-tll-
John-T (him) self-S Shakespearef0 (him)self-of friend-
eykey ilkecu-ki-lllscohahanta
PL-to read-NOM-0 likes
'John likes to reat.1 Shakespearefor his friends.'
b.john-nAn Shakespeare-lilchinku-t-U-eykey ilkecu-k1-3Al
cohahanta
'John likes to read Shakespearefor his friends.'
(5) a. John-nkliMary-l11ponay-nln taysin caki-ka ka-keysstako
John-T Mary-0 send-instead of (him)self-Sgo-will-QT
malhayesssta
said
'John said that he would gohimself instead of sending
Mary.'
ponay-mln taysin ka-keysstako . b. ?*john-rAn gary-lll
malhayessta
'John said that he would gohimself instead of sending
Mary.'
39G 187
silsu-ey tayhayese (6) a. john-nncaki-ka ecey cecili-n made-REL mistake-about John-T (him)self-S yesterday
sakwahayessta
apologized h.z made yesterday.' /John apologizedfor the mistake that tayhayese sakwahayessta b. John-nin eceycecill-n silsu-ey that he made yesterday., 'John apologizedfor the mistake 4-6 is correferentialwith The sub'ect of eachembedded sentence in The underlyingstructures that of the respectivematrix sentence. of 4-6 are shown as4/-61 below: chinku-tjel-eykey (41) John-nin jobn-kaShakespeare-lil John-ly john-of friend-PL7to John-T john-S Shakespeare-0
ilkecu-ki-lil cohahanta
read-NOM-0 likes ponay-nin taysin John-ka (51) John-nnjohn-ka Mary-lll send-instead of John-5 John-T John-S Mary-0
ka-keyss-tako malhayessta
go-will-QT said tayhayese (61) John-nln John-ka eceycecill-n silsu-ey
John=T John-S yesterdaymade-REL mistake-about
sakwahayessta
apologized applies to these REFLEXTTLEATION, whichis formulated below, of 4-6, respectively: structures and derivesthe (a) sentences
197 188
(7) REFLZXIVIZATION:
NP., Y], Z ===4> 1, cLki,
1 2 3 different from eLzh Sentences 4-6 representthree cases which are of the reflexive proncn: other in regard tothe deletability (2) the deletion is (1) the deletion isoblioatory in 4;
/ in 5; (3) thedeletion Ls lp to the of presupposition in 6. I will show belowhow the introductL.e apparely random in+-r, grammar enablesit to explain this will be provided by case. phenomenon. The explan,ition As in 4, thr- C- 0 1, wherethe deleton isolirl3orNi. obligatory in thefc deletion of reflexivepronoun is also
senence:
(8) a. T'John-nincaki-ka Mary-11sathavha-n
John-T (him)self-S Mary-0murdered-RE', suspicion-C;
patassta
received
'John was suspectedto have murdecedMary.,
patasst. b. John-nin Mary-lilsalhayha-n
'John was suspectedto have murderedMary.' is (:_v Compare 4 and 8with 9 and 10, wherethe reflexive pronoun
22p2aaly delotable: 4 Amy ilkey-toy-ki-lil palant.d (9) a. John-nisncaki-ka Snakespeare-lil read-become-NCX-0 hopes John-T (him) self-S Shakespeare-0 ad Shakespeare.' 'John hopes that hewill be appointedto 19E 189
(9) b. John-nin Shakespeare-111ilkey-toy-ki-111 palanta
'John hopes that he will beappointed to read Shakespeare.'
(10) a. John-nin caki-ka Mary-111salhayha-n yayki-lil hayessta
John-T (him)self-S Mary-0 murdered-REL story-0did
'John told the story of (his)haying killei Mary.'
b. John-nln Mary-111salhayha-n yayki-111 hayessta.
'John told the story of(his) having killed Mary.'
What could be the factorsthat mak( the deletion ofreflexive
pronouns obligatoryin 4 and 8, while optionallyleaving the
pronouns in 9 and 10?Could it be conditioned by thedifferent
10 have verb in the embedc:ed sentences? It could not, since C and
the same embedded verbsalhay hata. 'murder', but onlyin 8 is the
deletion obligatory. Could the differentcomplementizers cause the
di-:ference in deletability?No, for 4 and 9 haveidentical
complementizers, although theydiffer in the deletabilityof tne
reflexive pronouns. It does not follow, however,that we are
dealing with complete randomness. In fact, there is oneconsistent
factor that helps to explain thesituation. The difference in
deletability seems to comefrom the different n3tr1x verb: in
Korean (but not inEnglish), one cannot like (=cohahata) to read hue (. palata) if that reading done by others, although one can
that others may read. And on purely semanticgrounds, it would
hardly be possible for one tobe suspected of (vs hyemiy-111patta)
others' having committed murder, eventhough one can talk about
others. (, y./.....7..ilaki.jista) a murder's havingbeen committed by 199 190
In other words, 4 and 8 havematrix verbs which require embedded subjects to be correferential with matrixsubjects, and therefore there is no possibility of ambiguity asto the identity of the embedded subject when the embedded subjectis deleted. The matrix verbs of 9 and 10 do not, on the otherhand, require correferentiality betweea the subject of anembe sentence and that of a matrix sentence. A generalization can be made tentatively as follows: if the matrix verb is lexicallymarked for correferentiality between the subject of anembedded sentence and that of a matrix sentence, thenthe reflexive pronoun which isin the embedded sentence must be deleted.
Now, consider the following:
(11) a. 2sJohn-nin caki-kachuw-ta-ko malhayessta
Jbhn-T (him)self-S cold-QT said
said that he was cold.'
b. John-nitn chuwta-komalhayessta
'John said that he was cold.'
(12) a. 7*John-nin caki-kasuyeng-lil ha-ko siphta-ko
John-T (him)self-S swimming-0 do-NO:1 want-OT
malhayessta
said
1John said that he wantedto swim.'
b. John-nin suyeng-111ha-ko siphta-ko malhayessta
Ijohn said that he wantedto swim.' 200 191
(13) a. ?*John-nlncaki-ka enehak-lilcenkongha-keyEzta-ko
John-T (him)self-S 1iaguistics-0major-will-QT
malheyessta
said
'john said that hewould major inlinguistics.'
D. John-ninenehak-lilcenkongha-keyssta-ko malhayessta
'John said that hewould major inlinguistics.' sacu-ma-ko Mary-ewkey (14) a. ?*john-nincaki-ka chayk-1U Mary-to John-T (him)seli- book-0 buy-will-QT
yaksokhayessta
promised book for her.' 'John promised Marythat he would buy a
b. John-ninchayk-IA1 sacu-ma-koMary-eykey yaksokhayessta book for her.' 'John promised Marythat he would buy a morphemes like keyss or ma'future Lee (1970) hasascertained that chuw-ta 'feel cold' intention' and severalperceptional verbs like be used in embeddedsentences whose or siphta'feel like' can only subjects of thenext-higher subjects arecorreferential with the 1 possibility of ambiguity asto the sentences. There is no But in 11-14, identity of the embeddedsubject in suchinstances.
it is the verbof the embeddedclause that requires subject and the matrix correferentiality betweenthe embedded
subject, unlike 4 and8-10, where thematrix verb requires
correferentiality. with respect to 4and The generalizationthat has been made
8-10 must now beexpanded: 201 192
(15) If the subject of the embedded sentence is required to be
correferential with that of the matrix sentence either
because of the matrix verb or because of the embedded
verb, the embedded subject is obligatorily deleted.
Casta 2, where the deletion of a reflexive pronoun is blocked. It is easy to think of counterexamples like 5 and 16 to the expanded generalization given immediately above:
(16) John-nin caki-ka suyeng-111 ha-ko siphta-ko malhayessta
john-T (him) self-S swimming-0 do-NOM want-QT said
'John said that he wanted to swim himself.'
However, the Nills that are reflexivized in 5 and 16 areformally or
exhaustively used--they are key information in the embedded
sentences. That caki 'himself, in 5 and 16 is used formally can 2 be demonstrated by the direct narrative counterpartsto 5 and 16:
(51) a. John-nin !Mary-111 ponay-nth taysin kakeyssta,
John-T Mary-0 send-instead of I-S go-will
hako malhayessta
OT said
'John said, "I will go myself instead of sendingMary.",
b. John-nth 1Mary-litl ponay-nin taysinna-nin ka-keysstal
hako malhayessta
'John said, III will go instead of sending Mary.1 202 193
(161) a. john-nn suyeng-2411 ha-ko siphta, hako
John-T I-S swimming-0 do-NOM want QT
malhayessta
said
'John said, HI want to swimmyself."I
b. John-nin suyeng-1U ha-ko siphtal hako
malhayessta
'John said, HI want to swim.,
The direct quotations seemto be treated as independentof their matrjx sentences so far as thecycle is concerned: i.e. post-cyclic rules--which apply only after thefinal cycle--can apply in adirect quotai:ion before it itself isembedded.Therefore, if the underlined phrases na-ka in 5ta and 16I-1had been used materially,they could have been topicalized.However, the ungrammaticalityof 5,1) and the
lack of synonymy between16ta and 16Ib show that TOPZNis blocked.
Furthexmore, 5Ia and 16Ia can begrammatical only when the
underlined phrases are heavilyaccented.The fact that TOPZN is
blocked in 5Ia and 16ta and thefact that the NPIs must beaccented
in that eventualitydemonstrate that caki in 5 and16 stands for
formally used rPls. These reflexive pronouns thatstand for NPIs
used formally cannot bedeleted, as shown by theungrammaticality of
5b and the lack of synonymyof 16a with 12b. Since an NP used
formally carries the keyinformation of a sentence, itsdeletion will
necessarily result in aMbiguityas'to the identity of thereferent;
therefore, reflexive pronounsthat replace NPIs usedformally are ?Oa 194
. not deletable.Now I can expand the previous generalization to read
that only when the speaker presupposes that ambiguity does not
result can reflexive pronouns be deleted.
Case 3, where the deletion is up to the individual s aker's
judgment. By expanding the generalization to say that the reflexive
11111MI pronoun should be automatically deleted when the spee-aer does not
presuppose resultant ambiguity, I have shownthat the deletability
of a reflexive pronoun is actually conditioned by an individual
speaker's judgment.With the previous two cases, then, speakers of
Korean share presuppositions: In case 1, I have dealt with those
sentences concerning which speakers share a presuppositionthat
ambiguity is not possible; in case 2, we find sentencesconcerning
which speakers share the presupposition that ambiguityis
inescapable when an embedded NP is deleted. In case 3, I will deal
with those sentences about which speakers share nopresupposition
concerning any resultant ambiguity.Thus, consider the following
sentences:
(17) a. John-nin caki-ka Mary-kwa heyeci-te-n nal-1A1
John-T (him)self-S Mary-with separate-RET-REL day-0
kiekhakoissta
remembers
'John remembers the day when he left Mary.'
b. John-nin Mary-kwa heyeci-te-n nal-111 kiekhakoissta
'John remembers the day when he left Mary.,
Sentences 17a and 17b are equally grammaticaland are synonymous.
2Q4. 195
The deletion of the reflexive pronoun is optionalhere, not in the sense that the presence orabsence of it depends on a speaker's unconditioned choice, but in the sense that thedeletion is left to a speaker's judgment as tosituational need--if the speaker presupposes that ambiguity willbeset the identity of an embedded
NP when the reflexive pronoun is deleted,he will (mandatorily)
leave it; otherwise, he will obligatorilydelete the reflexive
pronoun.
In fact, this obligatory deletion ofredundant
information--which is dependent on the speaker'spresupposition
that reslAant ambiguity is unlikely--is notlimited to the deletion
of reflexive pronouns in embedded sentences. Thus, 18Bi is rather
awkward as a response to 18A, since theinformation carried by
the underlined na-ka (or -nin) isclearly redundant.Notice that
the choice between 18Bi and 18Bii isnot uncondi.r.ionally left to
the speaker: the choice of 18Bii is in factmandatory.
(18) Speaker A: (tangsin-nin) cikim etey Kaseyyo?
you-T now where go-QUES
'Where are you going?'
Speaker B: i. na-nin cikim hakkyo-eykapnita
I-T now school-to go
'I am going to school now.'
hakkyo-ey kapnita
,(I) go to school (now).1
Then the generalization on deletionof redundant information in
206 196
Korean can be given in the follyging manner:
(19) When and only when the speaker presupposes that resultant
ambiguity is not likely may reflexive pronouns
conveying redundant information be deleted.
G. Lakoff (1971:340) calls such a general constraint as 19 a global constraint. A global constraint conditions t.,e applicability of transformational rules. In my approach, then, the obligatoriness or optionality of a rule depends on global constraints in the grammar.
200 197
APPENDIX II: LIST OF RULES*
3.1 ADJECTIVE PREPOSING 4.2 CASE AGREEMENT 4.2 COMPLEMENTATION 2.14 CONJUNCTION REDUCTION ch. 4, rule 19 DATIVIZATION 3.12 DEFIN1TIZION 2.13 EQUI-NP DELETION ch. 31 rule V LEFT-SISTER-ADJUNCTION ch. 2, rule 96 NR (Adj-Pr) FORM ch. 3, rule 58 NR-FORM ch, 3, rule 7 NOT-TRANSPORTATION ch. 4, rule 18 OBJECT PARTICLEINSERTION 4.2 PARTICLE DELETION ch. 4, rules17-20 PARTICLE INSERTION 4.14 PREDICATE PREPOSITION ch. 4, rule 47 PREDICATE PROMOTION 3.11 PRONOMINALIZATION Appendix I, rule 7 REFLEXIVIZATION 3.1 and 3.21 RELATIVE CLAUSEREDUCTION 3.22 REPT-NP DELETION ch. 3, rule 58 RR-FORM 3.13 RR-PRONOMINALIZATION 3.12 S -DELETION 1
; 6 207 198
ch.4,rule :7 SUBJF,CT PARTIZLE LNSERTION
ch.4,rule 20 to-INSERTION
ch.4,rule 7 TOPICALIZATION
The ordering of thesesyles is so complicatedthat the detemination of theirprecise formulation isbeyond the scope ordering, of the presentlissertation.For part of the rule
see sect. 4.2. 191
APPENDIX III: LIST OF KOREAN PARTICLES
ki ci morphophonemic variant of ciman but cul NOM cungeyze from anon: ey inanimate nouns) ey directional (moving toward
ey(se) locative animate nouns) eykey directions, %moving toward
NOM
iy of
ka
k' NOM
ko OT
ko and
kwa with
1 REL with futuremeaning
la IMP
1&1 0
man only
myen if
REL
ney PL
ni when 209 200 nin nin DLM
(nin)tey and pakkey other than
pota than
puthe from
te RET
tey see pin tey
til PL
to even, too
to but
wi on
210 201
NOTES ON CHAPTERTWO
in 1 and 2 'Pre- and post-determinerCI,s are illustrated below, respectively:
(1) a. All (of) theboys
b. Many (ofthe) boys
(2) a. The manyboys
b. The five boys
1970:36-39. For furtherexplanation, see Carden VIII in his book arenot 2Carden,s argumentsIII, IV, V, and
applicable to Korean, asfar as I can see. embedded 3NT-verbs are thoseverbs that allowthe NEG in the and believe areNT-verbs, sentence to beraised. Such verbs as think below are synonymouswith and therefore the(a) and (b) sentences
each other: start until noon. (1) a. I think thatthe train will not start until noon. b. I do not thinkthat the train will date any girl. (2) a. I believethat John will not ever girl. b. I do nctbelieve that johnwill ever dateara 2b do not containNEG, the Although the embeddedsentences of lb and in an affirmativesentence) underlined morphemes(which cannot occur verbs have beenraised. For indicates that theNEG on the matrix and Baker 1970b. further discussion, seeLakoff 1970:145-65 4 I owe this term,delimiter, to Yang. negative verbs, see For an analysisof implicitly
Klima 1964b:294. 211 202
6For further discussion of definiteness and specificityof
NPIs, see Bailey1967:29-55.
7The correferential NP's maybe repeated when they arethe words thai: carry the newinformation of a sentence(formally-used
NP1s).
(1) Jcihn-ka cenhwa-111hayessko,JOhn-ka kkoch-lil sawassta
John-3telephone-0 did-andJohn-S flower-0 brought
'It was John who madethe phone call, and it wasJohn who
brought the flower.' 8 See sect. 3.31for the suggestedunderlying structures of
generic nouns.
9Some native speakersof English are dubiousabout the accept 1 acceptability of thissentence. But even such speakers
below:
(1) Those men were manyin number. For a 10A sentence like72b is here termed anintro-S.
detailed analysis of anintro-S, see sect. 3.11.
11The fact that all is not used as an overtpredicate seems
to be peculiar tocurrent English.
12_There are twodifferent kits: one is ademonstrative ki, close to the hearer but which is used to referto something that is the pro-form of arelative far apart from thespeaker, the other is 84a of this chapter. clause.For an example ofthe second kit see
13I claim in sect.3.11 that la below is aderived
structure which lbunderlies: 2411 203
(1) a. A certain mantried to conquer theworld.
(There was mani)-(manitried to conquer theworld)
the scope of a NEG, If the second conjunctof lb alone is within
2 will be derived:
(2) A certain mandid not try to conquerthe 1Norld.
If la is within the scopeof a NEG, 3 will bederived:
(3) No man tried to conquerthe world.
'There was not a mansuch that he wanted tc conquerthe
world.I
14This rule is actually a rule schemaconsisting of NR- This rule FORM, RELATIVE-CLAUSEREDUCTION, andADJECTIVE-PREPOSING. predicate. reads as NR-FORMapplying to a clausewith an adjective 15 The rule The verb deny, isderived from say that not. The derived not-HOPPING applies andraises not to thematrix verb. lexical spelling rule. structure pot say isrealized as deny by a which make implicitnegation. See Klima1964b:294 for other words verbs in an intro-S 16It appears that werarely use stative co-ccur or the unless such additionalphrases as TIME, LOCATION determine the stative verb carriessufficiem information to For this reason, identicy of a referentin a particularsituation. application of RR-FORMhere. I excluded 121in illustrating the
1 7Accidentallyidentical NPIs are thoseNP1s which have
result of different identical forms only onthe surface, as a Thus, flying transformational rulesfrom distinct deepstructures. accidentally identical planes in 1 andflying planes in 2 are
213 204 to each other.
(1) Flying planes can bedangerous (= 'To fly planes canbe
dangerous,).
(2) Flying planes can bedangerous (= 'Those planesthat are
flying can bedangerous1). ZO 5
NOTES ON CHAPTER THREE
1See sect. 2.2for two types ofadjectives.
2This also explains why B adjectivescannot be predicates
restrictive relative clauses of restrictiverelative clauses, for intro-S. will be shown laterin this chapterto be copies of an A for the sake of rhe discussion oftense will be ignored simplicity in expositionuntil Chapter V. 58 of this chapter. 4Forthe formulationof RR-FORM, see rule be, and 5RELATIVE-CLAUSE REDUCTIONdeletes relative pronoun +
ADJECTIVE-PREPOSING moves anadjective ahead of theNP it modifies,
as illustratedbelow:
(1) I bought aflower which is red.
(2) *1 bought aflower red.
(3) I bought a redflower.
6 here used in the sense The termscopying and chopping are A copyin2 type that Ross usesthem fo: reorderingtransformations. chopping one does not. See of movement ruleleaves a trace, but a
Ross 1967, sect.6.2.1, for details.
7 definite]. Cf exs. 291-31' Proper nouns canbe used as
of this chapterand the succeedingdiscussion. aFor different sources of definiteness, seeexs. 80-82.
9 used as a generic nounas 7: The noun manin this case is examples: horse, a horse,and horses are inthe following
(1) A horse is afaithful animal.
(2) The horse is afaithful animal.
2 1 S :ssal
(3) Horses are faithfulanimals. in Generics have beenclaimed to be definitein the sense suggested the present dissertation. article in An interestinganalysis of the definite connection with generic nounshas been suggestedby Vendler (1968:11-
:below: 25). Vendler argues that 2is derived from 2'
(2') The (animal thatis) horse is a faithfulanimal.
Vendler's work had beenbrought to my attentiononly after this
dissertation had beenfinished.
10For the definitionof representativeand partitive generic,------
see pp.27-28.
iSince it is mostly Type II intro-S thatis relevant in the
intro-S. discussion, from now onI will refer toit simply as an 12 imply the ordering ofbasic In Thompson1967, she seems to embedding from right-to- conjuncts bydistinguishing left-to-right 1970b, she seems tocontradict left embedding. But in Thompson
what she had claimedearlier. cannot 135ee sect. 2.2for detaileddiscussion of why Q's
have indefinitesubjects. 14 structure, in thetraditional I do not usethis term, deep (1967) have sense that means abase component. Lakoff and Ross I mean by this convincingly claimed thatthere is no suchlevel. surface structures term a logical orsemantic level underlying
generated from it. 15 concerning intro-Vs andzhe See sect. 3.11for details
respect to otherconjuncts. basic orderingof an intro-S with 218 207
16 See Appendix I fordetails concerning REFLEXIVIZATION and the deletability of reflexive pronouns.
17See note 6 of chapter 3for the definition of a chonp:Ing rule.
1 8SubjectNPIs modified by a RR-(B) cannotbe topicalized, as shown by theungrammaticality of 1 below:
(1) 7*ecey na-111 cuki-lyeha-n hansakong-nki spei-iessta
yesterday me-0 kill-tried-REL onesailor-T spy-was
IA sailor who tried to kill meyesterday was a spy.'
(2) ecey na -111 cuki-lyeha-nsakong-nin spei-iessta
IThe sailor who tried tokill me yesterday was a spy.I 19 Under the presentanalysis, step 1 is obligatoryipas shown
by the ungrammaticalityof la;
(1) a. *Seymour gave me aknife.; and I used a knife..
b. Seymour gave me a knifei;and I used the knifei.
The second occurrenceof any correferential NPhas to be definite
because of the obligatoriness ofstep 1.
20The deletion of S seems to me to be acomplete mistake. 1 21F or some apparentcounterexamples, see sect. 3.22.
22what I suggest as the underlyingstructure of 87a is
different from 87b. I suggest 1 below as theunderlying structure
of 87a; note the additionalbrackets in my suggested underlying
structure.
(1) (.-- a janitor...).( (I just met a janitor)-(A janitor
me this key)) 217 208
23,The interrelation among speaker, hearer,referent are intricately involved in the Korean honorific system.See Lee
1970:87-93 for details. 24 Particles are actually inserted by latertransformations, but added here to facilitate the interpretation.
25A stacked clause is similar toChomsky's nested construction.For further discussion of thismatter, see
Chomsky 1965:12.
26Somehow, speakers accept sentenceslike 1 (where a RR 4.n which the shared NP is the subject follows a RRin which the
shared NP is non-subject) more readily thansentences like 2 (where
the ordering of the RR's is opposite fromwhat is given above):
(1) There are many people that I know thatdo not know me%
(2) There are many people that do notknow me that I know.
Bailey 1969 has maae an interestingexperiment, in which a very neat
implicational scale of acceptability is revealed tobe conditioned by
different types of RR's, relative pronouns,etc. 27 Copula agreement and numberagreement seem to be later
rules, in spite of Lakoff-Ross's(1966) suggested rule ordering.
These authors claim that RELATIVE-CLAUSEFCRMATION comes later than
NUMBER AGREEMENT and COPULA AGREEMENT.But then, we could notexplafn
the agreement in a surface structurelike 105. 28 For the definition of thisterm, see pp. 27-28.
29For the definition of thisterm, see p. 88. 30 See sect. 5.3 for the suggestedanalysis of negation. 218 209
31F or explanationsof rules 129-131, seeCopi 1969:41-46 any property and 122-30.The Greek letterphi in rule 131 represents syrbo1. 2 1 below by the In other words,138 is converted into principle of distribution. to give flowery to (1) ((I bought(certain) flowerx)-(I want
Mary))-((I bought (certain)flowerx)-(flowerx is red)) major conjunct, the same Since the intro-S isgiven in the first
intro-S in the secondmajor conjunct isnecessarily presupposed Therefore, only RR-FORM(A) by the speaker tobe shared by hearers.
is possible inthe second majorconjunct.
219 210
NOTES ON CHAPTER FOUR
Ni'smodified 1Seesect. 2.2 for thederivation of indefinite by adjectives.
2See sect. 311 forthe definition ofintro-S. means 3We hame two distinctetten's in Korean: one whic dealing 'certain', amd the otherwhich means 'which'.We are here with etten in the firstreading. time) are 4The features E+ comb E+ loc3, dir], and C+ in the deep stActure. derived from predicatesof higher sentences shapes to thewfeatures, Then some low-levelrules give phonological For an ex npleof the which are realized asparticles on thesurface. into lower sent nces, process of loweringsuch higher predicates
see rule 19of this chapter. r+mentioned], is suggested by Dru ig(1968). 5This feature, in the dscourse in An NP is r.4. mentioned], if ithas been mentioned
which a givensentence is used.
6Generic nouns behavesimilarly and gettopicalize
becomes obligatory.See gnomic sentences,TOP= actually
for an example. 7 particles, seeang 1971. For detaileddiscussion of Korean int 8Since Korean is aSOV language,the first NP
and the second?IF is the NP, NP, Nr constructionis the subject,
object. 9 I use theEnglish For convenienceof exposition, sord *it yea an abstract 3which affect. But I assume thatthis represents 2 2 C) 211 determines the relationbetween a dative NP and anembedded sentence.
10Lakoff 1971 has madeit very clear thatpresuppositional
rules. information conditions theapplicability of transformational
But I have shown in thepresent dissertationthat this presuppositional description of information has to beincorporated in the structural a rule.
11The fact that bothcontrastive wa (or r-in) andexhaustive fact that both of them 22 (or ka) arenormally accented and the they are carry newinformation of sentencesmight suggest that further evidence related. But for the present, Ido not have any
to conclude thatthey are (or arenot) related. nln and 120ne might wish to arguethat since both thematic these two nkits are contrastive nIn needcontextual information, fact actually not distinctfrom each other. But in that case, the embedded sentence cannot that only contrastivenin can occur in an
be explained. 11Direct quotations behave differently fromother embedded
direct quotation whichis embedded sentences. TOPZN applies to a See pp. 192-93 for adiscussion into other performativesentences.
of this matter. 14_ the following martin (1964:408)claims that Korean has familiar, polite, six speech levels: plain, intimate, analyses of speech authoritative, anddeferential.For different
levels in Korean, seeLee 87-93, andSong 15-23. 15 NR (Adj-Pr, PREDICATE PREPOSING,1.oneexample of which is derivation of lh :CORM, is a ruleschemaconsist1ti4of NR-FORM (e.g. 22,1, 212
-from la) and RELATIVE-CLAUSE REDUCTION(e.g. derivation of lc from
lb). In case the predicate of thesecond conjunct is an adjective,
ADJECTIVE PREPOSING is includedin that schema. Otherwise, a
replacement occurs, as illustratedby the derivation of ld from lc
below:
(1) a. (x killed Mary).(x isJohn)
b. X, who is John, killedMary.
c. X, John,killed Mary.
d. John killed Mary.
16Contrastive-. nIn can mark thesubject NP's of these verbs.
17John-ly is derived from some deeper structure like
aphn-k ha-nln 'which Johndoes' by some sort ofrelative-clause
reduction rule. For some discussion ofRELATIVE-CLAUSE REDUCTION in
Korean, see Ree 1969,ch. 3. 18_ ror thederivation of dative NP's, seerule 19 in this
chapter.
19The morpheme series ninkes is a derived form of
la-ko ha-nln kes 'thingwhich they say...1 by deletingko ha A description, / gloss nin kes as NOM 'they sayy.For simplicityle
here.
0Forsome discussion of'inalienability*, see Fillmore
1968:61. 21F or examples ofagglutinative verbal, see Kuno1970:III 19-
21. 213
22_ror the specificcondition for deleting reflexive pronouns, see Appendix I.
23For S-pruning, seeRoss 1967:41-102.
223
t, 214
NOTES ON CHAPTER FIVE
1Fora discussionof NT verbs, see ch. 2, n.3.
2See ch. 4, n. 8.
3Lee might want to arguethat obligatory ha-DELETIONwill automatically delete hata. But in 5.22, I have shownthat ha-DEL is not motivated in Koreansyntax. Besides, if Korean has obligatory ha-DEL, the grammarcannot explain thedifference in grammaticality between thesentences in 1 and those in2 below:
(1) a. i kong-ka ttwi-ki-111hanta
this ball-S bounce-NOM-0does
'This ball does bounce.1
b. *i kong-kattwi-ki-ka hanta
this ball-S bounce-NOM-Sdoes
'This ball does bounce.'
c. *John-ninachim-puthe cenyek-kkaciyelsimhi kongpuha- study- John-T morning-from evening-tillhard
ki-111 hanta
NOM-0does
'John does studyhard from morning tillevening.1
(2) a. i kong-kdttwi-ki-to hanta
this ball-Sbounce-NOM-even does
'This ball evenbounces.,
b. i kong-kattwi-ki-n*n hanta
this ball-Sbounce-NOM-DLM does
'This ball does atleast bouncing.' 224 215
-to (2) c. John -nin achim-puthecenyek-kkaci yelsimhi kongpuha-ki
John-T morning-from evening-till hard study-NOM-even
hanta
does
'John even studies hardfrom morning till evening.'
For a detailedexplanation, see 5.22.
he following are someexamples of A and B verbs.
A verbs B verbs
plan, expect manage, remember, see fit decide, be venture, dare, have sense ç ready eager willing be enaull kind fortunate speaker's feeling or judgment A verbs are justthe expressions of Kajita (1967) calls semi- of the subject. This is similar to what
auxiliary verbs.
5Lee (65) has anotherrule which deletes ha,which he calls
ha-DEL-1. 137-39. 6Forthe discussion ofcontextual particles, see pp.
7If ADVB does not applyto 24b, COMPLEMENTATION
automatically applies tothe lowest sentenceof 124b, and 1 below
will be derived:
(1) John-ka kes-ka ppalita
john-S run-REL thing-S(= NOM) is-fast f lIt is fast that Johnruns.'
22 216
8TENSE-SPELLING adds a tensemorpheme (e.g. ess for thepast, nin for the present tenseof action verbs) to verbroots.
9Verbalizing hata, which wasdiscussed in p. 159, is one example of hatal.
10 hatais not in the deepstructure, As I have claimed, 2 but transformationallyinserted in sentences later. 217
NOTES ON APPENDIX I
1Perceptive verbs like chuwtaIfeel cold', silhta
'disliked', always have thefirst-person, singular subject ifin the present tense; if thetense of those verbs is past,there is no such restriction on the subject(e.g. ex. 3):
(1) na-nin chum-kachuko-siphta
I-T dance-S dance-want
'I feel like dancing.'
(2) *ne (or k)-ninchum-ka chuko-siphta
you (orhe) -T dance -S dance-want
'You (or he) feellike dancing.'
(3) ki-nin chum-kachuko-siphessta
IHe felt likedancing.I
This seems to be sobecause one can only be sureo..7 one's own
perceptions at a givenmoment. sect. 4.1. 2Forthe definition ofthe formal use of an NP, see
227 218
REFERENCES
Austin, J. L. 1955. How to do things withwords. Oxford: Oxford University Press. Universals in 3ach, Emmon. 1968. Nouns and noun phrases. linguistic theory, ed. byEmmon Bach and RobertT. Harms, Winston, Inc. 90-122. New York: Holt, Rinehart and Marshallese Bailey, C.-J. N. 1967. Transformational outline of Chicago: syntax. Ann Arbor: University Microfilms. University of Chicago.
1969. Marked and unmarkedorderings of two relative clauses modifying the sameantecedent. Working Papers in Linguistics,University of Hawaii,1:10.239-44. and result 1970a. The instrumental case, purpose clauses, and theunderlying sententialoperator cause. Working Papers inLinguistics, Universityof Hawaii, 2:9.155-60. Working 1970b. Trying to talk in the newparadigm. Papers in Linguistics,University of Hawaii,2:9.39-54. Linguistic Baker, C. L. 1970a. A note on scopeof quantifiers. Inquiry 1.138-40. 1.169-86. 1970b. Double negatives.Linguistic Inquiry the do-so rule. Papers Bouton,L. 1969. Identity constraints on in Linguistics1.231-47. requiring a non-sentential Brame,M. K. 1970. A derived nominal source. Linguistic Inquiry1.547-49. definiteness in relativesand Browne,Wayles. 1970a. Noun phrase Linguistic questions: evidence from Macedonian. Inquiry 1.267-70.
markers: interrogatives . 1970b. More on definiteness in Persian. Linguistic Inquiry1.359-63. English Studies40.163-69. Buyssens, E. 1959. Negative contexts. in 1970a. Logical, predicatesand idiolect variation Carden, Guy. Laboratory, English.Cambridge, Mass.: Aiken Computation Harvard University. (Report NSF-25.) 22tt 219
-Carden, Guy. 1970b. A problem with primacy. Linguistic Inquiry 1.523-33.
Casagrande, Jean. 1970. A case for globalderivational constraints. Papers in4Linguistics2.449-59.
Choi, H. B. 1961. Uli malponLAKorean grammar). Seoul: Chong um sa.
Chomsky, Noam. 1965. Aspects of the theory ofsyntax. Cambridge: M.I.T. Press. Readings in English . 1970. Remarks on nominalization. transformational grammar, ed.by R. Jakobson and P. S. Rosenbaum, 184-221. Waltham, Mass.: Ginn and Company.
1971. Deep structure, surfacestructure, and semantic interpretation. (To appear in Aninterdisciplinary reader in philosophy,linguistics, and psychology,ed. by D, D. Steinberg andL. A. Jakobovits.Cambridge, Mass.: Cambridge UniversityPress.) The Copi, Irving M. 1967/1969. Symbolic logic. New York: Macmillan Company. Language 46.33-41. Dahl,Osten. 1970. Some notes onindefinites. clauses in Drubig, Bernhard. 1968. Some remarks on relative English. Journal of English as aSecond Language 3.23-40.
Fillmore, Charles J. 1963.The position of embedding transformations in a grammar.Word 19.208-31.
1968. The case for case. Universals in linguistic theory, ed. by Emmon Bachand Robert T. Harmc,1-88. New York:Holt, Rinehart andWinston, Inc. Readings 1970. The grammar of hittingand breaking. Jakobson and in English transformationalgrammar, ed. by R. P. S. Rosenbaum,120-33. Waltham, Mass.:Ginn and Company.
Harris, Z. S. 1964. Discourse analysis. The structure of language, ed. by JerryFodor and Jerrold Katz,355-83. Englewood Cliffs, N.J.:Prentice-Hall, Inc. degrees of definiteness Hetzron, R. 1970. Nonverbal sentences and in Hungarian. Language 46.899-927. relative clause reduction. Huddleston, Rodney. 1971. A problem in Linguistic Inquiry2.115-16.. 229 220
Jackendoff, Ray S. 1968a. Quantifier in English.Foundations of Language 4.422-42.
1968b. On some incorrect notionsabout quantifiers and negations. (To appear in Studies intransformational grammar and relatedtopics, principalinvestigator: S. J. Keyser. (ASCR1-68-0032.1Waltham, Mass.: Brandeis University.) determiners. . 1968c. Speculations on presentences and Unpublished.
1969. An interpretative theoryof negation. Foundations of Language5.218-41.
1971. Gapping and related rules.Linguistic Inquiry 2.21-36.
Jacobs, Roderick and PeterRosenbaum. 1970. Readings in English transformational grammar.Waltham, Mass.:Ginn and Company. Forge Jespersen, Otto. 1964. Essentials of English grammar. Village, Mass.: University of Alabama Press. relative 3osephs, Lewis S. 1971. Tense and aspect in Japanese clauses. (To appear in Language,1971.) study of Kajita, Masaru. 1967. A generative-transformational semi-auxiliaries in present-dayAmerican English.Tokyo: Sanseido Company LTD.
Kartunnen, Lauri. 1967. The 2dentity of nounphrases. (Publication no. P-3756.) Santa Monica, California: The RAND Corporation.
1968. What do referential indicesrefer to? MMENNs(Publication no. P-3854.) Santa Monica, California: The RAND Corporation.
Unpublished. . 1969. A-verbs and B-verbs.
1971. What makes definite nounphrases definite? (Publication no. P-3871.) Santa Monica, Califorhial The RAND Corporation.
Kim, S.-H. P. 1967. A transformationalanalysis of negation in Ann Arbor: Korean. Unpublished Ph.L. dissertation. University of Michigan. 2Sd 221
'Hiparsky, Paul and CarolKiparsky. 1970. Fact. Progress in linguistics: a collectionof papers, ed. by Manfred Bierwisch and Karl E.Heidolph. The Hague: Mouton and Company. The structure of Klima, Edward S. 1964. Negation in English. language,ed. by Jerry A.Fodor and Jerrold J. Katz, Prentice-Hall, Inc. 246-323. Englewood Cliffs, N.J.: Journal of Koutsoudas, Andreas. 1968. On wh-words inEnglish. Linguistics 4.267-73. (Report NSF-27.) Kuno, Susumu. 1970. Notes on Japanese grammar. Cambridge, Mass.:Aiken ComputationLaboratory, Harvard University. and certain related Kuroda, S.-Y. 1968. English relativization (Reprinted in Modern problems. Language 44.244-66. studies in English, ed.by David Reibel andSandford Schane, 1969.) syntactic irregularity. Lakoff, George. 1965. On the nature of Aiken Computation (Report NSF-16.) Cambridge, Mass.: Laboratory, HarvardUniversity.
verbs in English.. . 1966a. Stative adjectives and (Report NSF-17.) Cambridge, Mass.:Aiken Computation Laboratory, HarvardUniversity.
(Report NSF-17.) Cambridge, . 1966b. A note on negation. University. Mass.: Aiken ComputationLaboratory, Harvard
1966c.Deep and surface grammar. (Duplicated.) Linguistics Club. Bloomington: Indiana University
Instrumental adverbs andthe concept of deep . 1968a. structure.Foundations of Language4.4-29. Counterparts, or the problemof reference in . 1968b. transformational grammar. (Duplicated.) Bloomington: Indiana UniversityLinguistics Club.
constraints. Papers from the . 1969. On derivational fifth regional meetingof the Chicagolinguistic society, Department of ed. by Binnick etal., 117-39.Chicago: Linguistics, Universityof Chicago.
logic. Ann Arbor, . 1970a. Linguistics and natural the University ofMichigan. Michigan: Phonetics Laboratory,
3 3 1 222
he -Lakoff, George. 1970b. Pronominalization, negation, andtational analysis of adverbs.Readings in English transfol grammar, ed. byR.Jakobson and P. S. Rosenbaum,10 Waltham, Mass.: Ginn and Company.
'negation, conjInction, . 1970c. Repartee, or a reply to and quantifiers'.Foundations of Language6.189-422-
descriptively inadequat = . 1970d. An example of a interpretative theory. Linguistic Inquiry1.539-42-
1970e. Global rules.Language 46.627-39. An On generativesemantics. (To appear ir 1971a. and interdisciplinary readerin philosophy,linguistic3, wits. psychology, ed. by D.D. Steinberg and L.A. Jakol Cadaridge, Mass.:Cambridge UniversityPress.) well-formecless. 1971b. Presupposition and relative interdisciplinary reader inphilcsophy, (ro appear in An and linguistics, andpsychology ed. by D. D.Steinber CambridgeV Liversity L. A. Jakobovits.Cambridge, Mass.: Press.) of cransformational and John R. Ross. 1966. On the ordering , Indiana 'olversity rules. (Duplicated.) Bloomington: Linguistics Club. 'Juplicated.) 1967. Is deep structurenecessary? Liner_stics Club. Bloomington: Indiana University source. 1970. A derived nomiriAl 4uiring a sentential
Latin complementation. Lakoff, Robin. 1968. Abstract syntax and Cambridge, Mass.: M.I.T. Press. there can't be anysome-any rule. . 1969. Some reasons why Language45.608-15. bub's aboutconjunction. . 1970a. If's, and's, and Unpublished. participants. 1970b. Tense and itsrelation to Language 46.838-49. to' construction. Langendaen, D. T. 1970. The 'can't seem Linguistic Inquiry1.25-36. Seoul:Language Lee, M. S. 1969.Nominalizations in Korean. University. Research Center,Seoul National
2.?e. 223
Pan Korea Lee, H. B. 1970. A study of Koreansyntax. Seoul: Book Corporation.
Lindholm, J. M. 1969. Negative-raising and sentence regional meeting pronominalization.Papers from the fifth Binnick et al., of the Chicagolincuistic society, ed. by University of 148-58.Chicav): Department of Linguistics, Chicago. and Korea. Language Martin, Samuel. 1964. Speech levels in Japan in culture and society: a reader inlinguistics and anthropology, ed. by DellHymes, 407-15. New York:Harper & Row, Publishers. of languages. McCawley, James. 1967. Meaning and the description Kotoba no Uchu 2:9-11. Tokyo: Ename of publisher not available2. Universals 1968a. The role of semanticsin a grammar. and Robert T. in linguistic theory,ed. by Emmon Bach Winston, Inc. Harms, 125-69. New York: Holt, Rinehart and component of a 1968b. Concerning the base transformational grammar. Foundations of Language 4.243-369. Language 46.286-99. 1970a. English as a VSO language. Readings in 1970b.Where do noun phrases comefrom? R. Jakobson and English transformationalgrammar, ed. by Ginn and Company. P. S. Rosenbaum,166-83.Waltham, Mass.: of presuppositionin Morgan, Jerry L. 1969. On the treatment fifth regional transformational grammar.Papers from the by Binnick meeting of the Chicagolinguistic society, ed. Linguistics, et al., 167-77. Chicago: Department of University of Chicago. to the complex NP Oh, Choon-Kyu. 1970. Spurious counterexamples constraint including a'variably crazy rule'in Korean of:Hawaii, syntax. Working Papers inLinguistics, University 2:9.105-24. and quantifiers: Partee, B. H. 1968. Negation, conjunction, 4.153-65. syntax vs. semantics. Foundations of Language Progress in Perlmutter, D. M. 1970. On the article inEnglish. of papers, ed. by M.Bierwisch and linguistics: a collection Mouton and Company. K. E. Heidolph.The Hague: 224
-Perlmutter, D. M. 1968. Deep and surfacestructure constraints in syntax. Unpublished Ph.D.dissertation. Cambridge, Mass.: M.I.T.
1970. Surface structureconstraints in syntax. Linguistic Inquiry1.187-256. Relative clauses withsplit , and John Ross. 1970. antecedents. Linguistic Inquiry 1.350. in English.Modern Postal, Paul. 1969. On so-called 'pronouns' studies in English, ed.by David Reibel andSandford Prentice-Hall, Inc. Schane, 201-24. Englewood Cliffs, N.J.: Unpublished Ree, Joe J. 1969. Some aspects of Koreansyntax. Ph.D. dissertation. Bloomington: Indiana University. Modern stUdies Reibel, David andSandford Schane (eds.). 1969. Prentice-Hall, Inc. in English.Englewood Cliffs, N.J.: constraint on negationin Rivero, M. L. 1970. A surface structure Spanish. Language 46.640-66. principles of English Rosenbaum, Peter S. 1967. Phrase structure 3.103-18. complex sentenceformation. Journal of Linguistics governing deletion inEnglish . 1970. A principle aNOIMNM transform- sentential complementation. Readings in English ational grammar, ed.by R. A. Jacobs andP. S. Rosenbaum, 20-29.Waltham, Mass.: Ginn and Company.
Ross, John R. 1963. Negation. Unpublished. clauses. (Report 1966. Relativization in extraposed NSF-17.) Cambridge, Mass.: Aiken Computation Laboratory, Harvard University. in syntax. Unpublished . 1967. Constraints on variables M.I.T. Ph.D. dissertation. Cambridge, Mass.:
Auxiliaries as main verbs. (Studies in . 1969a. Philosophical Linguisticsseries, 1)1 77-102.
Papers from thefifth regional . 1969b. Guess who. meeting of the Chicagolinguistic society, ed.1.)y Binnick et al., 252-86.Chicago:Department of LinguibLics, University of Chicago. Readings in English . 1970. On declarativesentences. P. S. transformational grammar,ed. by R.S. Jacobs and Ginn and Company. Rosenbaum, 222-72.'Waltham, Mass.: 225
Rutherford, W. E. 1970. Some observationsconcerning subordinate clauses in English. Language 46.97-115. Papers in Sadock, Jerrold M. 1969a. Super-hypersentences. Linguistics 1.1-15. Papers in Linguistics1.283-370. . 1969b. Hypersentences. system of old saxon. Schotta, S. G. 1969. The relative pronoun Papers in Linguistics1.182-93. modifiers in English. Smith, Carlota. 1961. A class of complex Language 36.342-65. clauses in a generative 1964. Determiners and relative (Reprinted in grammar ofEnglish. Language 40.37-52. Reibel and . Modern studies inEnglish, ed. by David Sandford Schane,1969.) rules in Korean. 1967. Some transformational Song, S. C. Indiana Unpublished Ph.D.dissertation. Bloomington: University. transformational Stockwell, R. P.,et al. 1968. Integration of theories on Englishsyntax. Unpublished.
Thompson, S. A. 1967. Relative clauses andconjunctions. Working Papers inLinguistics, The OhioState University, 1.80-99. and constraints on . 1970a. Relative clause structures Linguistics, types of complexsentences. Working Papers in The Ohio StateUniversity, 6.20-40. Working 1970b. The deep structureof relative clauses. University, 6.41-58. Papers in Linguistics,The Ohio State
nominalizations. (Papers on Vendler, Zeno. 1968. Adjectives and 5.) The Hague:Mouton and Company. 0 formal linguistics, markers in Korean. A study ofdelimiters and case Yang, I. S. 1971. University of Unpublished Ph.D.dissertation. Honolulu: Hawaii. (In preparation.) in syntax. Arnold M. Jr. 1968. Naturalness arguments Zwicky, Chicago Papers from thefourth regionalmeeting of the al., 94-102. linguistic society,ed. by Bailey et of Chicago. Chicago:Department ofLinguistics, University 235