1

1

•

i

Llnsulst1c Theory and Second Language Acquisition: the Acquisition

of Engli sh Reft exi ves by Nat i ve Speakers of Japanese

by

-

='~-:r-~--'~~-~- r __ ~ ~ Maklko Hirakawa - - -

-- ~; L ~ =-~.: _ ~ ~ -~ .. ~ -- - -::

::!: =-

-_;-_~--'-~";~C:- ~-__ :_-~~~ __

~ ~_-t-_

~- _~"_r~~;_'::._

, c

A Thesis Submitted to the Faculty of Graduate Studies and Research

in Partial Fulfillment of the Requirements for the Degree of

Master of Arts

,--~

~ -"",-

~ .... --~- :;:':-

, r~l

Department of Linguistics McGill University Montreal. Quebec

@Makiko Hirakawa April. 1989

1

1

l

,

Abstract

Linguistic Theory and Second Language Acquisi tion: the Acquisi tion

of English Reflexives by Native Speakers of Japanese

1 n the framework of generati ve grammar, Uni verf;al Grammar has

been posited as a language acquisition device with which a11 human

beings are innately endowed and which consists of invariant

principles and variant parameters. Assuming that Uni versaI Grammar

is involved in first language acquisi tion, this i:hesis investigates

how Universal Grammar can be related to second language acquisition,

as we II as the ro 1 e of the fi rs t 1 anguage in 1 earn i ng a second language.

An experi mental study was conducted to exam i ne how na t i ve

speakers of Japanese acquire syntactic propert i es of English

reflexives. In particular, the effects of the Governing Category

Parameter and the Proper Antecedent Parame ter were studied. English

and Japanese differ with respect to their values of these parameters.

The resul ts support a transfer hypothesis; i.e. the parameter

setting of the first language was transferred in acqui ring the L2 setting, leading to transfer errors. However, 1 suggest that

parame ter resetting is possible in the long run, which argues for

the operation of Universal Grammar in second language acquisition.

Makiko Hirakawa

M.A.

- i -

Department of Linguistics

McG ill Uni versity

1

1

Résumé

Linguistic Theory and Second Language Acquisition: the Acquisi tion

of Engl ish Reflexi ves by Nati ve Speakers o\f Japanese

Dans le cadre de la Grammaire Générative, on considère la

Grammaire Uni verselle comme un ensemble de princi '.~'es invariants

associés à un nombre limi té de paramètres. Selon c\':tte approche,

la Grammaire Uni verselle joue le role de procedure d':a\pprentissage

du language. En supposant que la Grammaire Uni verselle est

opérative dans l'apprentissage de la langue mater:nelle, nous

explorons dans ce mémoire le role de la Grammaire Universelle dans

l'apprentissage de la langue seconde.

Une étude experimentale a été conduite afin de déterminer

comment les locuteurs natifs du japonnais apprennent les propriétés

syntaxiques des pronoms reflexifs de l'anglais. Spécifiquement,

nous avons étudiés l~ffet du paramètre de la Catégorie Gouvernante

et celui de l'Antécédent Propre. L'anglais et le japonnais

different crucialement relativement à la valeur assignée à ces

paramètres.

Les resultats de cette élude supportent l'hypothèse du

transfert i.e. que la valeur du paramètre determiné pour la langue

maternelle est transférée dans l'apprentissage de la langue seconde,

conduisant ainsi à la production d'erreurs. Cependant, je suggère

que la valeur du paramètre peut être réétablie dans un délais plus

long. Ceci suggère que la Grammaire Universelle joue tout de même

un role dans l'apprentissage de la langue seconde.

Makiko Hi rakawa

M.A.

-ii-

Department of Linguistics

McGill University

1

T

Acknow 1 edgcments

1 would like to express my sinccre gratitude to aIl those who

have helped me with my studies and !Tl} life here at McGi11.

First and foremost. 1 am forever graleful ta my thesis advisor,

Prof. Lydia Whi te. Her knowledge and her enthusiasm have gi ven me

insight into the nature of language acquisi tian study. Her comments,

suggestions and criticisms have becn invaluablc to me in the course

of writing lhis thesis. and 1 am (~specially indebled to her for the

time and cncrgy she dcvoted tu helping me polish the final draft.

1 would like to express my apprcciation la the faculty of the

Department of Linguistics at McGill; in particular, Prof. Lisa

Travis, who first awoke my intcrest in syntax and made suggestions

whenever 1 asked for them: Prof. Mark Baker. from whom l learned how

to think about and resolve problems of language; and Prof. Nicole

Dom i ngue, for her warmhearted enrouragement.

1 have greatly benefi ted from my contact wi th the students of

the Linguistics Department in rnany ways. through thcir help and

discussions, both in and out of class. 1 would like to express

special thanks to Jane Macaulay for careful proofreading of this

thesis; and ta José Bonneau for preparing the French abstract.

1 have also benefited from the helpful comments made by Prof.

Mamoru Saito and the members of the Mita Psycholinguistic Seminar

on this thesis experi ment.

1

1

1 am grateful to Laurraine Coffin for her help in the

statistical analyses of the experiment.

1 would like to thank Minoru Okamoto at Meikei High School.

Prof. Takeshi Teruki at the College of Foreign Studies Yokohama, and

Marisa Persechino at LaSalle College for allowing me ta conduct the

experiments with their students. 1 thank aIl the students who

participated in the experiment for their cooporation.

My graduate studies in Canada were made possible by a Rotary

International Foundation Scholarship and Teaching Assistantships in

the McGill Linguistics Department and in the Center for- East Asian

Studies, for which 1 am grateful.

1 also wish ta thank my friends who have made my studies and

stay in Montreal wonderful; especially, Françoise, Eva, Jane, José,

Marika, Masako, Sachiko, Shanley, and Sumi.

Many thanks la the members of my family in Japan, Wakiko

Hirakawa, Yasuyuki Ki noshi ta, and Toki Ki noshi ta, for their love and

unders tand i ng.

Finally, ta Yahiro, for his endless encouragement across the

sea.

-iv-

1

1

,

Abstracl

Résumé

Acknowledgemenls

Table of Conlpnls

Cahpter 1 INTROOllCT ION

fable of Conlt'nls

Chapter 2 UN IVERSAL GRMMAR AND LANGUAGE ACQU IS Il ION

2.1. Innale Prüp('rt.i(~-:; of Mmd

2.2. The l.oglcal ProblPm of Languaw' Acquisi lion

2.3. Pnnciplcs ~md Paramptf'rs of l'G

2. 4. UG i n &'Cood Languagp Acqu i s i li on

2.·1. 1. Di fferC'ncps Md Simlla,· illf'S bC'tween LI

2.4.2. The Loglcal ProblC'm of L2 Acquisi lion

2.4.3. UG and L2 acqulsi lion

2.4.3.1. The 'antl-UG' hypothesis

2.4.3.2. lh:> 'U:;' hypolhesis

2.4.3.3. The 1 transfer' hyçothesis

Footnoles ta Chapter 2

Chapter 3 PARMll:iERS OF UG

3. 1. Anaphors

3.2. Binding fneory

3.3. The Goveming Category Parameter

3.4. Type (e) language - Japanese

éUld L2 Acquisi tian

3.4. 1. PrQJertÏe:; of Japanese Reflexive 'Zibun t

3.4.2. Goveming Calegory for Zibun

3.5. The Proper Antecedent Parame ter

Foolnoles ta Chapter 3

-v-

i

11

iii

v

1

4

4

6

9

11

12

14

15

16

18

20

Z3

24 :E

:E 30

32

32

34

36

39

1 Chapter 4 LEARNAB IL lTY AND THE SUBSET PH INC IPLE

4.1. The Subset Principle

40 41

4.2. The Subsel Principle and L2 Acquisi tian 46

4.2.1. Sludies by White (1987) and lobl (1988) 46

4.2.2. The Sludy by Fiœr and Broselow (1986) 48

4.2.3. The Sludy by Thomas (1988) 51

4.3. Preference vs Grammar 55

4.4. Hypolheses for the Ex-perimenlal Sludy 58

Footnoles ta Chapler 4 62

Chapter 5 EXPER 1 MENT AL STUDY 63

5.1. Rational behind the Sludy and Two Hypolheses 63

5.2. Melhod 64-5.2.1. Subjects 64-5.2.2. Materials 67

5.2.2.1. Syntax Test 68

5.2.2.2. UG Test 72

5.3. Results 76

5.3.1. RffiUltS of lhe Experimental Groups 76

5.3.2. ExJ:erimentai Group vs Engi ish ContraIs and JapaJ1e3e 82

Conlrais

5.3.2.1. TYÇ-e A 'w Type D

5.3.2.2. Type E

5.3.3. Preference 's pragmatic Factors

5.4. Discussion

Foolnoles la Cl'lélpler 5

Chapter 6 CONCLUSION

Appetldix A

Appendix B

Appendix C

References

-vi-

84-86 87

91

98

101

104-105 1~

108

1

1

Chapter

INTRODUCTION

The main objecti vc of this lhesis is ta lest experimentally how

and ta \\lIat exlent nativp. speakers of .Japanese learning English

acqllire properties of tnglish rcfle,ivc pronouns. and to investigate

the raie of the first language (LI) in learning the second language

(L2). L2 acquisition often refcrs to a ~ituation \\'hÎ~re a language

is learned in a community or country "here thaL language is used as

a native language by sorne of t.he pf~oplf~ living there; for example.

French learned by English speaking people in the province of Quebec.

Canada. or English learned by non-native speakers in the United

States. Ho\\ever. t.he L2 may be a foreign language. a language

learned where residents do not use that language, as exernplified by

English learned by nati ve speakers of Japanese in Japan. As a basic

assumption, 1 will adopt the idea that L2 acquisition includes

foreign language learni ng, since both processes i nvol ve a learner

internalizing a grammatical system that accounts for the target

language.

ln this thesis, theories developed ln L1 acqllisi tian research

-1-

1

vdl1 be applip.d to L2 acquisi tion. have chosen to examine certain

abstract properties of the synlax of English. basing my arguments on

assumptions made \\i thin gcnerativc grammélr, particularly Government

and Binding (GB) theory (Chomsky, 19~1). GB theory assumes that

there is a Universal Grammar (UG) \\hich human beings are innately

endowed \\i th and \\hich constrains language acquisi tion. Assuming UG

constrains LI acquisition, my first concern lS whether L2

acquisition IS also conslrained by L:Ci. If L2 l(~arners have access

to UG, do they recapi lulate the COllrSt~ of LI acquisi tion? In order

to investigale thcse questions. 1 condurled an aquisitional study on

native speakers of .Japanesc I\ho \H~re learning English as a foreign

language through thcir formai p.ducation in .Japan.

In Chapter 2. discuss UG, tlH~ rationdle behind it. its

contenl, and i ts relevance to L2 (}cquisi tion. The concepts of

principles and parameters are also explained. Chapter 3 deals with

the syntax of English and Japanp.se \\ith respect to the properties of

reflexive pronouns. TI\o paramctcrs, the Governing Category

Parameter and the Propf'r An tecedent Parame ter (Wexl er and Manzi ni.

1987; Manzi ni and Wexler. 1987), \\ i Il be introduced. The

acquisition of these parameters will be examined in the experiment.

The aim of Chaptcr 4 is to explore learnability problems in LI

acquisi tion and in L2 acquisi tion. The Subset Principle (Berwick,

1985; Wexlcr and Manzini. 1987: Manzini and Wexler, 1937). which

has bccn proposed t 0 so 1 ve stllh a prohlem, IS considered.

Furthermore, invcstigate predict ions \\hich follow the Subset

2 --

1

Principle as well as alternative hypotheses in case the principle

does not operate. Alternatives include a transfer of the LI

parameter setting and a transfer of LI preferences. The latter 1S

motivated, as it seems likely that native speakers have a preference

for one possibility in cases where the grammar allows severa!.

The rest of the lhesis reports an experimental study of the L2

acquisition of English reflexives and investigales how L2 learners

acquire the conditions on reflexives when their LI yields a superset

of the L2 along a markedness hierarchy. The results obtained here

turned out ta be problematic for the Subset Principle. They support

the LI transfer hypolhesis. However, there are subjects who

correctIy reset to the L2 val ue of the parameter. Thus, 1 will

suggest a few stages by which L2 Iearners gradually retrcat from the

mosl marked setting, their LI value, to the least marked setting,

their L2 val ue. The resul ts support the idea that UG is acti vated

in L2 acquisition and that, although a transfer of the LI occurs,

resetting ta the correct L2 value is eventually possible.

-3-

a

1

1

Chaptel'" ? ..

UNIVERSAL GRAMMAR AND LANGUAGE ACQUISITION

Language acquisi tion is unique to human beings. Moreover. it

is accompli shed by children in a comparati vely short period. even

though the language system acquired by children is very complex. How

is such language acquisi tion possible" Investigation of the nature

of language acquisi tion is assumed by generati ve grammarians to

reveal properties of natural language and the human mind. This

chapter discusses general issues "hich are crucial for the theory of

Uni versaI Grammar (UG) and examines to what extent these issues can

be extended to L2 acqu i s i tian.

2.1. Innate Properties of Mind

In the framework of generati ve grammar, it has been assumed

that sorne innate properties of mind are invoived in language

acquisition; and it is UG that has been posited as an innately given

language acquisition device.

Chomsky (1986a) states that generative grammar is:

-4-

1

1

,

concerned with those aspects of form and meanlng that are

determined by the "language faculty," \\hich is understood

to be a particular component of the human mind. The nature

of this facul ty is the subject matter of a general theory

of linguistic structure that aims to discover the

framework of pri nciples and el ements corn mon to aUainable

human 1 anguages; UG may be regarded as a

characterization of the genetically determined language

facul ty, One may think of this facul ty as a "language

acquisition device," an innate component of the human mind

that yields a particular language through interaction wi th

presented experience, a device that converts experience

into a system of knowledge attained: knowledge of one or

another language, (1986a:3)

Postulation of such an endowment stems from what is called the

'logical problem' of language acquisi tion or the 'poverty of the

stimulus', Our knowledge of language IS complex, abstract, and

subtle. People share such knowledge in a language community;

nevertheless, such knowledge cannot be assumed to be acqui red by

children only through experience, or the input provided to them,

-5--

1

1

2.2. The Logical Problem of Language Acquisi tion

Native speakers have internalized a complex grammar \\hich

underlies their use of language. The logical problem concerns how

people come to attain this system. When ~e compare the input data

available ta a child and the grammar which he or she eventually

attains, there seems to be a discrepancy between the two. This

mismatch has been discussed in the literature by Chomsky (1982,

1986. 1988), Hornstein and Lightfoot (1982), and White (in

preparation). Briefly, the problem is that a child does not seem to

receive aIl the necessary information about the target language. if

input data is assumed to be the only source available ta a child who

is constructing the grammar, \\here this input data inc1udes

caretaker speech, a11 the utterances available to the child, and the

child's experience with the outer world.

Let us consider sorne illustrations of the mismatch between the

input and the grarnmar of English. The sentences in (1) i 11 ustrate

wh-movernent in English: (lb) and (Id) are grammatical wh-questions

but (If) is ungrammatical l•

-6-

1 (Ua. Mary bought a piano.

b. What, did [,PMary buy t, ?]

c. John heard that Mary bought a piano.

d. What, did [",John hear [ t, that [IP Mary bought h.?]]]

t 1 t ______________ ~

e. John heard the news that Mary bOllght a piano.

f.*What, did[,PJohn hear[NPthe news [:~ . .!-that[lpMary bought c?]]]]

t * 1 t'-______ _

These sentences show that a simple generalization from observed

facts could lead the child astray. That is, (If) wOllld be a

logically possible question equivalent of (le), if the child simply

induces a general rule of wh-question formation from sentences

(la }-(Id). However, (If) is ungrammatical. Thus, wh-questions

cannot al ways be formed by replaci ng a noun phrase (NP) by wh-phrase

and putting it in a sentence ini tial posi tion, and inverting the

subject and an an auxiliary element. This is evidence for the fact

that linguistic competence is more complicated and abstract than we

might assume.

(If) has been explained as a violation of a · subjacency

constraint'. InformaI1y, the subjacency constraint states that no

NP can be moved by crossing more than one bounding node. In the

case of English, NP and IP nodes are bounding nodes; thus, in (If)

a wh-phrase has crossed two bounding nodes, the NP and the

-7-

1

f

t

matrix IP. sa that the sentence is ruled out.

~ati ve speakers of English recogni Le that sentences like (If)

are incorrect in spi te of the f act tha t they have never been taught

their ungrammaticality. Thus. \\e may conclude that the input

underdetermines the grammar of language; l.e. the input is far less

complex than the fundamental properties of the attained grammar.

This underdetermination must be compensated for; linguists argue

that innate principles of ~G gi ve the child the abstract knOl\ledge

whi ch i s not present in the i npu t.

It has also been argued that no negative evidence lS available

to a child (Brown and Hanlon. 19(0) 2. Negative evidence lS

information that tells a child \\hat is impossible in the target

1 anguage. How can a ch 11 d di scover if he or she has made an

incorrect hypothesis about the language \\Îthout being told that

something is not allo\\ed? That is. there is a possibil i ty that the

child overgeneralizes certain linguistic facts and that makes a

wrong hypothesis; for example. sentences like (If) as possible

English wh-questions. However. such overgeneralization leads to a

serious problem as it means that the child cornes ta possess a

grammar which is incompatible wi th an adul t grammar. gi ven that the

child will never be corrected on these mistakes.

The input is fini te as \\el1; in contrast. the output is

potentially infini te. Chi Idren can crea te new sentences. going far

beyond the input data that they ha\>p. actually heard. The input

varies from child to child. as every child goes through a different

-- 8 --

1

.,

linguistic experience. However. the child ends wi th the same

grammar as others do, and becomes able to deal wi th infini te range

of sentences.

These observations have led generati ve grammarians to postulate

UG. lnnate principles and parameters are assumed to guide the child

during acquisi tion, explaining how the child arri ves at knowledge

which goes far beyond the input, and how the child cornes ta know

about unglAammaticali ty wi thout the benefi t of negati ve evidence.

2.3. Principles and Parameters of UG

As there are numerous languages in the world, and every child

becomes a natl ve speaker of some language, UG must be able to

explain acquisi tion of any language. On the one hand, UG must be

restrictive enough to explain what is not allowed ln existing

languages. On the other hand, UG must be flexible enough to explain

the di versi ty of human 1 anguages.

In order to account for the acquisition of any natural

language, UG is assumed to be composed of two kinds of constraints:

principles, which are true for aIl languages where applicable, and

parameters, whi ch have different setti ngs. One example of a

principle is the subjacency constraint. As illustraied in section

2.2., subjacency prevents wh-movement across more than one bounding

node. What counts as a boundi ng node 1 s assumed to be

parameterized. In English, IP and NP nodes are bounding nodes; ln

-9-

contrast. CP and NP nodes are bounding nodes in ltalian as shown in

(2). Examples are from van RiemsdUk and Williams (1986: 71-72),

(2) a. Tuo domando [cpche storie!

your brother to whom myself 1 -ask which stories

[ IPabbiano raLcontato ~ ~ era molto preoccupato

they-have told was very worried

'Your brother. to whom 1 wonder which stories they have told,

was very worri ed:

b.*Tuo fratello. (cpa cui 1 [IPtemo [NPla possibilità [cpt 1 che

your brother to whom I-fear the possibility that

[ IPabbiano raccontato tutto t 1

they-have told everything

'Your brother to whom 1 fear the possibility that they have

told everything

(2a) is grammatical. although a cui 'to whom' has crossed IP and CP

nodes, since only one bounding node, CP, has been crossed. (2b) is

ungrammatical, as a cui has crossed two bounding nodes, CP and NP.

Thus. various subtle differences ln wh-movement possibilities ln

different languages are accounted for by one parameter of UG.

UG is based on a number of fundamental principles that

sharply restrict the c1ass of attainable grammars and

-} 0-

1

•

narro\\ ly constrai n thei r form, but wi th parameters that

have to be fi xed by exper i ence. If these parame ters are

embedded in a theory of CG that is sufficiently rich in

structure, then the languages that are determined by

fixing their values one \\'ay or another will appear ta be

quite diverse, since the consequences of one set of

choices may be very different from the consequences of

another set: yet at the same time, 1 imi ted evidence.

just sufficient to fi x the parameters of UG, will

determine a grammar that may be very intricate and will

in general lack grounding in experience in the sense of

inductive basis. (Chomsky, 1981: 3-4)

With principles and parameters, therefore, UG both limits and ,

explains existing languages. The assumption that the language

learner sets the values of parameters on the basis of evidence from

the input will be crucial for the experimental study in this

thesis.

2.4. UG in Second Language Acquisition

1 have mentioned that i t is not only plausible but also

necessary to assume UG in LI acquisition. That is, the human mind

is equipped \\ith a language acquisition dpvice which constrains and

guides LI aC'quisition. Given thélt UG contrais LI acquisition, then

1 l

(

(

the question arises as ta whether UG mediates L2 acquisition as

well. 1s there a logical problem of L2 acquisition? Ooes UG

mediate L2 acquisition exactly as it does LI acquisition? What IS

the role of the LI when a person learns the L2? In this section, 1

will considet" the theory of UG in relation t0 L2 acquisition

studil;!:..

2.4.1. Differences and Similarities between LI and L2 Acquisition

When we consi der L2 acquisition, there are various aspects

which differ from LI acquisition. First of aIl, L2 learners already

know their nati ve language. And although they have mastered their

LI wi thout any difficul ty, the y do not necessarily master the L2

successfully. Secondly, the age of learners 1S different. Children

start learning their LI from birth, while L2 learners' starting

points vary. Thirdly, the learning si tuati on may be different,

ln that L2 learning often takes place in a classroom situation while

LI learning is al ways in a natural setting. A child most often

starts his or her linguistic experience with caretakers, while a

L2 learner does sa with a language teacher. Consequently, there

IS a signlficant difference between LI and L2 acquisition with

respect to the amount of exposure ta the input. Furthermore, as L2

learners are cogni tively advanced, they may learn a language

consciously; ver·y young children are, in contJ"ast, quite unaware of

the fact lhal they are learning the language. Lastly, affective

-] 2-

1 factors, such as motivation, personality, apptitude, etc., seem ta

influence L2 acquisi tion but not LI llcquisi tion.

Ho\\ever, these various fact.ors mel:' affect the rate but not the

route of L2 acquisition (Ellis, 1936). It has been noted that the

process of acquiring L2 is rule-governed; l.e. various stages that

learners go through are not random at ail but sho\\ the systematicity

of learners internalizing the target language grammar (Selinker.

1972: Adjémian, 1976). Both Ll and L2 acquisition are processes of

acquiring language \\Ïth learners sdf-selectinp; the input data.

That is, L2 as \\ell as Ll learm~rs select. the input where

appropriate and utili/c it as evidence for constructing certain

rules of grammar, but ignore the inpuL \\here it is not relevanl. In

this sense, they are actively, rather than passively, involved in

the task of language acquisi tion.

To summarize. 1 \\ould like to focus on the route or process of

L2 acquisi tian. Abstracting a\\ay the various factors which affect

the L2 learner, L2 acquisi tian has this ln common wi th LI

acquisition: it IS a systematic attempt by a learner ta aCQuire a

grammar of the target language. Thi s thesi s focuses on the question

of what the L2 learner cornes to know about subtle properties of

syntax ln the target language and whcther t.he L2 lcarner's

i nterlanguage appears to obey the pri nci pl es and parameters of UG.

1

1

2.4.2. The 'LogicaI Problem' of L2 Acquisition

Does a L2 learner face the same logical problem as a LI learner

does? The most serious problem of underdetermination in L1

acquisition is clairned to hold in L2 acquisition as weIl (White,

1985b; Cook, 1985),

Let us consider the following set of sentences (examples are

from White, 1985b:30).

( 3) a. The book is duB.

b. Is the book du 11 ?

c. The book which is on the shelf is dull.

d. *ls the book which on the shelf i s duB?

e. Is the book which is on the shelf dull?

These sentences are related to the principle of 'structure-

dependency'. If L2 learners, on the basis of (3a) and (3b),

hypothesize that subject-auxiliary inversion takes place ln

question formation and that the left-most verb in the sentence must

be moved ta the initial position, we might expect them to produce

ungrammat ical sentence like (3d). This would be consistent wi th the

input. Such an error would show that learners treat sentences

linearly, ignoring the fact that the sentence has a hierarchical

structure. It has been reported, in fact, that LI acquirers do not

make such errors, suggesting that they are constrained by

-1 4-

1 structure-dependency (Crai n and Nakayama, 1987).

The information that (3d) is ungrammatical and (3e) is

grammatical is not the kind of information lhat is usually presented

ln the classroom situation. It appears that the input available to

L2 learners underdetermines the grammar of the target language just

as is the case in LI acquisi tion. In the case where the Ll and the

L2 have different properties. learners face a logical problem of L2

acquisition: neither their LI nor their language teachers give them

sufficient information about lhe 1:2.

If L2 learners attain complex knowledge of the target language,

which is underdetermined by t.he L:2 input available to them. it

suggests that UG must still be involved in L2 acquisition to solve

the logical problem of L2 acquisi tion.

2.4.3. UG and L2 Acquisition

Currently, there are three standpoints regarding the operation

of UG in L2 acquisition: UGj~ __ n~J9n~~I'" __ ()P!!raJjve (Bley-Vroman, in

press; Clahsen and Muysken. 1986 1; Schachter. 1988a.b), UG is

available but LI interferes (Flynn. 1987; Phinney, 1987: White,

1985a, 1936a, 1987b), ~G ___ operates exactly _tJ.!~~ame~y as in LI

(Mazurker\'i i ch, 1988). 1 wi Il go through mai n arguments of these

three posi tions below.

-} 5-

r

,

1

,

2.4.3.1. The • Anti -UG' Hypothesis

People arguing for this hypothesis state that UG is not

operative in L2 acquisition, on the basis of the following observed

facts in L2 aequisi tion: lack of eompleteness. variation in degree

in suceess, no equipotentiality (a ehild is equipot.-::ntial for any

natural language given normal input, while an aùult is not).

previous knowledge (L2 learners have al ready mastered thei r L 1).

existanee of fossilization in L2 (LI learners always master their

LI. while L2 learners tend to permanently stabilize at certain stage

which is short of suceess), indetcrminate intuitions (even very

advaneed L2 learners seem to lack cl~ar grammaticality jlldgements).

plausibili ty of help of instruction. practiee. and negative

evidenee (which L2 learners seem ta be indebted tO, but which has

nothing ta do with LI acquisition) etc. (Bley-Vroman. in press;

Schachter, 1988a)

Bley-- Vroman ( in press) argues that L2 acquisition is

fundamentally different from LI acquisi tion in t.hat L2 acquisi tian

is guided by nati ve language knowlerlge and general problem sol ving

systems. while LI acquisition is mediated by UG and domain-specifie

learning procedures. He claims that the L2 learner has access to UG

only through the LI grammar; and where LI knowledge is not

applicable, general learni ng strateg ies wi Il be lIsed; consequently,

adults show limited but variable ability in judgements where

principles of UG are involved.

- 1 6

1

,. ,

.1

1

As far as the post-puberty adul t L2 learner is concerne d,

Schacter (1988a) claims as fo11ows:

Grammatical competence IS presumably involved in the

production and comprehension of linguistic messages, but

is far from being the sole mechanism invol ved in their

transmission. Communicati ve capabili ty, al though making

use of grammatical competence, clearly invol ves many other

eapabili ties as weIl. Many second language speakers are

able to eommunieate effeetively in spite of obviously

defeeti ve grammatical competence. (1988a:224)

She emphasizes that even advanced learners of English do not

have · fully formed determiner systems, aspeetual systems,' and do

not produce 'movement transformations, particularly raising, clefts,

pseudoelefts, topicalizing rules, adverb movement rules' and

concludes lhat this lack of completeness can be accounted for by the

non-operation of UG.

However, 1 would like to argue that she is mistaken in judging

L2 learners' competence by thei r performance. doubt that

sentences with such movement transformation rules are used equally,

both quantati vely and qualitati vely, among nati ve speakers.

believe, for example, that use of honorifie expressions ln Japanese

native speakers varies from person lo person. It IS a basie

assumption of the UG framework thal knowledge of language is

-11-

1

1

1

internally represented in a grammar, and that it is not al\\ays fully

reflected in performance.

2.4.3.2. The 'UG' Hypothesis

To show that UG is accessible to L2 learners, we need ta have a

situation \\here sorne principle is observed in the L2 but not in the

learners' Ll. and \\here nothing regarding that principle has been

taught to the learners: this eliminates the possibility of the

learners basing their intuitions on their Ll or \\hat they have

been taught about the L2. If L2 learners observe the constraints of

UG in these situations, it suggests that CG opera tes in L2

acquisi tion. and that the anti -UG hypothesis is untenable.

Ritchie (1978) exarnined such a situation and found that the

'right roof constraint' \\as observed in 20 advanced Japanese

learners of English. The right roof ,-onstraint places limitations

on rightv,ard movement: an element cannot be moved to the right out

of the sentence in y.,hich the element originated. The examples in

(4) illustrate that (4d) is ruled out by this constraint (sentences

are from Ritchie, 1978:37L

(4)a. That John had Lelt surprised Mary,

b. l t surprised Mary that John had lelt.

c. [sThat it surprised Mary that John had Lelt] amused Alice.

d.*[sThat it surprised Mary] amused Alice that John had lelt.

-18-

1 Ritchie's subjects \\ere aSKed tü .illdgc pdirs nf sentences like

those above. Overall resul ts show t !l,II tilt; constrai nt \\()S obSt'r\'t'd

by Japanese learn{'rs; i.e .. tht'Y could dhtinp,ui~h bd\\('('n

permissible and impermissihle righh\,jrd movement in ~nglish. As

Japanese is not a language \\hich ohst~rvt'S such a conslraint (since

i t has no rightward movement). and as the constrainl. was not taught.

the only possible explanation of the result lS that UG was

accessible to these learners.

Another study by Ol.su and ~aoi (1~36) examines the principle of

structure-dependency in Il .Japane~e junior hip,h sellnol st udents.

These learners \\ere élsked to transform ~.nglish tlt:cldrative Stmtences

into interrogatives. They \\cre giv<m sent.ences with a relative

clause such as (5e) below. (Examplcs h/wc bCfm illustrat.f:d in (3)

in section 2.4.2 .. and are repeated in (S).)

(5) a. The book lS dull.

b. Is the book dul} ?

c. The book which 1S on the shelf is dull.

d. *ls the book \\hich on the shelf i s dull?

e. Is the book \\hich is on the shelf dull?

Japanese question formation fJrovides no clue as to how

questions should be formed in tn1,) ish. 1 f the } earner produces

questions like (Sc) rathf'r than (:=jd), if ~1JP,?csts t.hal tll~ or she

observes t.he principk of stnHt.lJn·dl'[)(\ndfm(y. If. in contrasl,

1

r

1

1

1

the learner produces sentences like (=1d), it could be concluded that

he or she does not observe the principle.

Seven of Otsu and ~aoïs subjects produced responses 'hhich

observed structure-dependency whi le three produced responses which

are different from (5e) but explainable by UG: and only one subject

failed ta observe t.he principle (only one third of her responses

observed the constrai nt). This study is also a case where L1 is not

relevant for the principle examined: therefore, UG must have been

acti vated.

2.4.3.3. The 'Transfer' Hypothesis

The transfer hypothesis argues for accessi bil i ty of UG in L2

acquisi tian, but daims that L2 acquisi tian will be influenced by

the LI. Many studies have exarnined cases where two languages differ

in their val ues for sorne parame ter. 1 n such cases, the LI val ue is

likely to be adopted (White, 1985a, 1986a, 1987b) or ta have other

effects· (Flynn, 1987). The studies which examine a parameter

setting model in L2 acquisition suggest that L2 learners do not set

pararneters as children do; their LI setting interferes initially so

that they choose the wrong value. If L2 learners have access to UG

directly, then no transfer errors should be found; if, however, L2

learners are affected by their LI parameter setting, then transfer

errors are to be expected. T1H~se transfer errors are not

i ncompat i ble wi th the theory of CG: moreover, they may be predi cted

- 2 0--

1 by variation in parcllIlPt ers Ilf lli.

White (1()8ba) ,lrgUt'S for \ht> Ir,lf1"ttT In'polhesis, Sht~ (',\drnilwd

'hhether or not paramt'Irll' \ariat ion h\'h\t't'n tilt' Ll ,Inti tht' 12 \\llllld

cause lrouhlt> for 12 lt'drncr~, kadinp, to tr,lnsf{\r ('rrors. lh(~

paramelcr cxam i ned \\ilS tht' SO (,Il kd 'pro drop' p,lrdfllt'lt>r.

Languages 1 i ke Spanish élnd 1!.1I i an cIrC pro drop Lmguagcs in t h,ll

subject pronouns can bt' omillcd. Tht~sC l(mguag('s (llso allo\\' Iree

inversion of subjed and vprb in th'el.lr,üive scnt{~m'(~s (namely, V S

arder) and the (~xtract ion of suhjl'ds uut of clauses \'\ith a

complementi/er (namely, lital traC't ' (iltt~r violations). '-.np,lish. on

the 0 ther hand. i s no t il pro drop LlIlgUilgc ~() t ha t • pro drop'. • V S

arder' and . that tra(,t~ filler violai ions' drp no 1. obs('rvt~d. ~clt.ive

speakers of Spanish and l'n~n('h \l'clrning l'nv,lish wen' \l'sl.cd on Illl~se

three prOpt~rt.if~s of rnp,lish by h\o tdc..;ks: grammat jcality ,iudg\'mcnts

and question-formation. \\hllf' inv(~sligdl('d whcthcr or Iloi d Sp,lI1ish

group, \'vhose LI had lhp pro drop set 1 inp; of Ihe par,lInct('r, would

have morE: difficult.y in sctt.in~ th(! ('orrt:!ct value for "np,lish l\tdn

a Frenrh ?,roup \\l!osc '1 h<ld il non pro drop seU i n~ Il ke \- np, 1 i ~h.

Although bath groups did weIl on 'V S ord~r' senl(~n('{~s, \ he Spanish

group made sip;nificant.\y more mi~tilk(!s ln judging 'pro drop'

sentences and in c<lrryÎng 0111 th(~ qtJ(~!-,t Ion formation tilsk for

'that. trace filler' !:)(~ntcn(!s, 'Ihc!-,f! rcsults slJgP;(~st th,li parts of

the LI par ametcr wcn~ crtrr i (!d ow!r t 0 t h(! 1.2,

To sum up Ihis dldpt(~r, 1 \\ill "<!IJpt UI!! followinl1, asslImptions.

2 1

•

t

1

,

1. UG operates in LI acquisition and L2 acquisi tian. UG consists

of principles and parameters, bath of \\hich restrict existing

languages as weIl as allawing them ta vary. UG is assumed ta be a

language faculty wi th which all human beings are innately endowed;

it maps the input data an ta the target grammar and guides language

acquisi tian.

Il. Althaugh UG is available ta L2 learners. L2 acquisition may be

interfered with by the Li. when two languages differ in values for a

certain parame ter of UG.

- - -L ~

_- _-=-r~ ~-~~- ~-

--~ ~ _______ ~ ~.:\ _ -" _ ~ r --'- -~- -~-

~ ~ ~ -~

-. -0..,-

- ::; - -- -

. --

-- 2 2-

J

,

Footnotcs ta Chapter 2:

1. Rcgarding the structure of clauses. 1 n:-;surnc in this thc~;is th.1l

sentcnce (S) is the maximal projection of INFl. (I). henc!' II': S IS lh('

maximal projection of Complcmcnti7cr (e). henccCP ([ollowing Chomsky. Il)~()

b }. San cl 1 P m a y b eus c d 1 nt cre han g C il b 1 Y. d S m il y S .In cl CP. Il. i n die il tes

that a sentence is ungnlmmatlcal or impossÎ bIc

2 . The r car c pc 0 pic who arr. u c for tb (' él V ail d Il i 1 i t Y (J flle & d t 1 ve (' v ide nec

Far c x a m pic, Il i r s h - Pas c k . 1 r c i man and S clin cid c r man (1 C) 8 1) cl ,li m,on III c

basis of pt'oduction data from ~O mothcrs .lnd thelr 2 to 1) veilr old chlldrcn.

th atm 0 the r s are mal' e i n cl i n c d t 0 r c p e a t U Il P, l' il m m ,d i C il 1 sen tell (' C :,. t h d Il

grammatical ones, which provldes subtlc ('uc~ [or the ('hild t 0 le!.!rn dllout

grammaticality. Howevcr. thcir result onlv hold::-. for 2-Y<'dr old chtldren:

m 0 r e a v cr. the f a ct th il t the l' i s mat cr n aIr cf 0 r m III il t 1 CI n 0 f r. r ri m m il li c ,d

sentcnces as weIl leads to the question of how a child Cdn distln&uish

betwecn two kinds o[ reformultaions.

3 . C 1 a h sen il n d Muy s ken (1 98 b ) l il k e il rat h crs t r 0 n fi .! n t i - U(j JI 0 ~d t.i 0 n .

Although Blcy-·Vroman and Schachtcr allow for the p(J:,,~.,ÎbtlI1 y of 1.2 ledrncr::,'

ace css i bill t Y t u lJ G t h r 0 u g h 1 Il ci ,. 1 1. C 1 a la ~'> c n rl n d M Il V ~, k (' n c 1 il 1 m 1 Il cd lH; i!>

not availablc at aIl tu adult !cdl-n('r',. ba~·,('d on thr' C(Jmpclr1~.(}n of (i('rm,1O

word order acquisition of LI and 1.2 l('arncr:,: 1.21<'drn(:r:, crc,lte rule:. wblcb

arc n 0 t d cf i n a b 1 e i n 1 i n gUI S t 1 (' t It cor Y Il Il 1 d (! l' 1 V (' d f l' Cl m t Il (' 1.10 11. Il tl !'. C

proccssing strategie::, and Reneral problf'm :.olving !.trdt('?,lc:; Du PI(':.~;j:.,.

Solin. Travis and WhIte (JC)87l how('vcr Y.IV(' d dlff(:r<'nt an.s1v:.i:. of C\.ll!:wn

and Muyskcn'~, data. cxplaÎnlnp, (j('rman 1.2 ,H qlli:;!llon in tl'rm:; of (Hi

~_ Flynn (1CJ37l cLlIm~ that Ih(' dlffcn;nt :.(·Itln~~:, lB LI ,111<112 cau:,c

difficulty and dclay ln êlcquirJnr. the L2 :.(,t llnp, on tltc l,rl!.I:, of hr'r :;1 11<1 y

of two grollp: .. Sp;mn.h and .Jap,lnc:.c, cH IjlllrInp, anrlptlC/rtl r<'ldll()n:,hlJ)~, tri

f:np,lÎsh ln rontra~;t tv WhIte. Flynn doc:·, not d:.~.lIm(' that the' 1) :,('Itlnp, Ît.;

initial}y adoptt'd by 1.2 lcarner:,

1

,

Chapter 3

PARAMETERS OF UG

In this chapter, 1 would like ta look into GB theory with

respect to the behavior of reflexive pronouns in English and

Japanese; In other words, what princi pIes of UG constrain

reflexives, and how the principles dîffer between the two languages.

A reflexi ve lS an anaphor which reqUlres an antecedent

syntactically and semantically. The Binding theory states the

essential conditions on anaphors, pronominals and lexical noun

phrases. l will focus on the constraints on reflexives, such as

himself or herself in English, and zibun in Japanese, as they are

the main concern of the experimental study in this thesis.

1 will present twa parameters: the Governing Category Parameter

and the Proper Antecedent Parameter (Wexler and Manzini, 1987;

Manzini and Wexler, 1987). It has been suggested that English and

Japanese differ wi th respect to the val ues of these parameters. 1

will show how and to what extent two languages may differ.

-24-

1

,

3.1. Anaphor~

Anaphors are '~Ps Ihal rt"'quin' .10 ,mlt\('t\(kal. In petrI iC'lll.tr.

reflexives, reciprocals. ,lI1d obi ig,lfory conlrol PBO ,ln' el/1clphor"s:

(van Riemsdijk & W i Iii ams 198(,: 111 )

lack their o\\n ref(\rf:nce rmd. ('on~t'qlJ('ntly. thf'Y dt\pf~nd for th(~ir

reference on an anlec('dfml \\11 ich l' ('owmands t t\t~m.

The following ~cntcn('es show , hat t h(~ ,m,lphor rc('(~ives i ts

ref erence through the an t cCf!dt!nt. i .t!.. i t. is coreft!rcnt. wi t.h the

under li ned NP.

(6) a. . John. likes himself, . ---- ----~--

b. John. and ~_art J 1 ikc each ot.l!f~r IJ' ----

c. John. want.s PRO, to go to the movic.

3.2. Binding Theory

Binding Theory states the followinp, princip\(!s:

(7) A. An anaphor must. be bound in i t s p,OVf~rn i np, ca t.f!p,ory.

B. A pronom i na 1 must b(! [rel! ln i t s p;OVf~rn i np, ( ah!p,ory.

C. A lexical NP must. be fn!(' (!vcry\\hen!.

( Van HÎf!msdijk ilnd Will i ams, 1 (JR7 :277)

1 will briefly rp.vi(!\\ t.h(! ddini t ion~ \\hi('h iln! (nwial for

2 ;)

1

1

Principle A. Defini tions are ci ted from van Riemsdijk and Williams,

1936 (henceforth, R & W). First, notions of c-command and binding

are defined as follows:

(8) C-command: A c-commands B if and only if the first

branching node dominating A also dominates B, and A

does not i tse 1f dom i nate B. (R & W:142)

Thus, in the follo\\ing configuration, ~ and Q c--command ~ and every

node dom i nated by c. e c-commands f, g and h, but not band c. - - - --

(9) a

(adapted from R & W:142)

(10) Binding: X binds Y if X and Y are coindexed and X

c-commands Y. (R & W:266)

- 2 fi --

1

1

,

(II) Governing category: a is il governing category for X

if and only if a is the minimal catcRory cont.aininp, X.

a governor of X, and a SlJBJECT é1ccessi hl<> t 0 X.

a. SlJBJECT: Th(' SUB.JECl of a (' 1 alise i s 1 Mm,. SI if

there lS one, o1.herwise !NP"Sl or INP"NPI (where

[X,Y] means uthe X immediat.ely dorninalpd hy Y"

b. accessibility: a is accessibIP to B if and only jf

a c-commands fi and the ass i gnmen t 0 f I.he inde>: of

a to P does nol lead to il violation of the

i-within-j Condition.

c. i-within-i Condition: *[7 ... 8 ... J7 where T and 8

have the same index

The sentences in (12) and (1'3) providc CXillIlp}CS of how I.hesc

principles work. The st.ruct.ure of t.he (a) !;(~nl(~IH'(, i~, givcn in t.he

corresponding (b) sentence (exampl(~s élrc froH! II li W:27(,).

(12) a. *John expects that himsel f will win.

b. [!l.NP, AGR, V [li :-thal. [..,2 him~cJf, AGIL VPIIII

(13) a. John expecl~ pictures of hims!!lf 10 be on sale.

b. [s,NP, AGR, V [\2[N!'2 picturf!!; of hirn:;elf.1 t.o VPJ]

ln {12bL the governor of him~,>clf lS Mm, in S~. which is a!:-.o a

- 27

: ..

f

SUBJECT accessible to himself. Thus. S2 is the minimal category

containing himself. its governor, and a SUBJECT. As himself is not

bound inside S2. the sentence is ungrammatical. 1 n (13b) . the

governor of himself is of. and the closest SUBJECT is AGR, of SI as

there is no SUBJECT in NP2. The governing category is the whole

sentence. St. Since himself is bound by John in Stl the sentence is

grammatical.

Given these defini tions. Principle A predicts the

grammaticali ty and ungrammaticali ty of the following sentences:

(14) a. [John, likes himsclf,l.

b. John, thinks that ~Bi1l1 likes himself IJ.

c. *John, thinks that ~Bi1l1 likes himsel f. J.

d. *John, thinks that lb i msel f , likes Bill J]'

e. John, told Bi III [PRO J to wash himself 1]'

In each case. square brackets sho\\ the governi ng ca tegory. 1 n (a),

himself is bOllnd by John in its governing category. In (b), the

governing category for himself is the lo~cr clause where it is bound

by Bill: while in (c), himself is bound by .Jo~. which is outside of

the governing category: therefore, himself fails to be bound in i ts

governing category and the sentence is ungrammatical on that

Interpretation. When himself is the subject of the embedded tensed

clause as in (d), it always fails to be bound in the governing

category; thus, the sentence is ruled out. The verb. tell. is

1 called a control v(~rb: Bill is an oh.h,t of IIIt' m.tin (1.111\1' ,md

there is PRO in the t~mbt~dded sllbjt'l'I position. (PIW IS ('Ll~sifi('d

as [+anapl!orÎc. 'pronominal:: ('onst'ljlll'nlly. il 1'.1 IH10 1 ht' ROvt'rncdJ

In (e), PHO is conlrolled by Bill. tht' ('omplt'mtml of IIlt' mdlll v(~rh.

and himself lakes .1S i ls ant.en~(knt. t.he cont.rollinR ~P. Bill.

through PRO.

The sen tences in (1 S) have lhe pronoun hi m i ns lt'.ld 0 f ri

ref 1 ex i Vf~. Pr i nc i pie B S t.a tes Uld t. a pronoull mus t b(~ f rc(~ in ils

governing category: lhus. t.h(~ followin!1, indt'\(~S are pn~dÎC'led.

( 15) a. [John, likcs himk;.

b. John, thinks lhat [Bil L 1 i k(~s him, il.

c. .John, lhinks that lhe, , k 1 i b~s Bi 1 L.l

d. .John, to 1 d Bi III [PRO I t.o wdsh him, k ) •

(An index k rcfcrs t.o an ex t.rascn Len t.i il 1 NP. )

As long as a pronoun 1 s no t. hound in 1 hl! ROVf!rn i ng ca h!l1,ory , lhn

sentence is !1,rammilt ica 1.

Following Principln C. a lexical NP like .John and Bill in (14)

and (15) ahOVf! musl be [rc(! in thc: 'hhole ~enl.(!n('e. lhi:-; mf!an:-;

neilher .John or Bill can bl! c commandl'd hy Ih(! ('oindc!xpd (!)Pmf'nt in

the sen tcn('(~.

This is a brid n!vÎc'h of Bindinp, !tH'ory. If IlH~S(! principlr!s

an! con:-;trdÎnls \\hich form prlrt of \,(i, tt,Py ffill:-..t !told tnlf' cro~~

linp,uisticrdly, II()\H!VI~r. it has bf!l!n ~)IJp.RI'~t('d 11Irt! Ihe p'ovf!rning

? !,

t

1

categories are subject to parametric variation (Wexler and Manzini

1987; Manzini and Wexler. 1987). The next section deais with such

a pararneter and di fferences 'l.hich resul t across languages.

3.3. The Governing Catp.gory Parameter

Languages may differ with respect to the domains in which an

anaphor must have i ts antecedent. Wexler and Manzi ni (1987) have

proposed that the governing category for anaphors and pronominals is

parameteri zed.

(16) a is a governing category for ~ iff

a is the minimal category which cantains Band

a. has a subject. or

b. has an INFL, or

c. has a TNS. or

d. has an indicative TNS, or

e. has a root TNS

( Wexler and Manzini 1987:53)

This is a slightly modified verSiOn of (11), but both essentially

mean the same thing. Each condition (a)-(e) corresponds to the way

the parameter rnay be realized in different languages. We have seen

above that English is a type (a) languaRc. Ho\\ever. other languages

may chaose another value from among the five possibilities. It has

- :~ ()

f

1

been suggested tha t 1 tali an j s a lype ( b): Russ i an. a type ( c ) ;

Icelandic, a type (d): Japanese and Korean, type (e) (Wcxler and

Manzini. 1987: Finer and Broselow, 1986).

According to the Governing Cat.t!gory Parameter, languages may

differ wi th respect to how far away lhe anleredenl can be f rom the

reflexive. as shown in (17),

(7) a. Keith, said that Ronnie) requires t.hat. Bill k persuade

Charlie 1 to consider Mickm fond of himself ..

(Finer and Broselow, 1986)

b. Language type Possible anf.ecedenl(s) for himself

a. Mick

b. Mick, Charlie

c. Mick, Charlie, Bi Il

d. Mick, Charlie. Bi 1 I. Honnie

e. Mick. Char~l i e. Bill. Honnie, Kei t.h

It can be said that a type (a) language is lhe mosl. restriclive

language in that il allows only lhe NP closf!st 1.0 Lhe rcflexivc to

be i ts anteceden t. Th i s type i s reprcscn t cd by ",ng 1 i sh. On lhe

other hand. a type (e ) 1 anguage, slJch as Japancsc. i s t.he 1 (!é}S t.

restrictive language in that il allows any NP (acllla!ly émy sllbject.

NP. as will be explain(!d lal.c!r) 10 be t.he émt.c('(-:dcnl. of lhn

reflexi ve. As ~an be sp.en from the I.abl e, thcse fj va t.ypes of

-:~ ]-

1

1

languages forrn a hierarchy regardi ng the possible antecedent (s) of

the reflexive. A type (e) language încludes every NP that is

allowed in other types of languages thai rlppear above (e) in Table

07b). In other 'Nards, it includes aIl the possibilities of other

languages; thus, it is considered ta be the most inclusive. A type

(d) language includes (c),(b),(a): a type (c) includes (b) and (a):

a type (b) includes (a), and, lastly, a type (a) includes no other

types.

3.4. Type (e) Language - Japanese

The experirnental study reported in this thesÎs examines the

si tuation in which the subjects' LI is Japanese and the L2 is

English. Since LI transf er is a factor that wi 11 be examined

further on, 1 \\ould like ta briefly reVlew sorne properties of the

Japanese reflexi ve pronoun zi bun.

3.4.1. Properties of the Japanese Reflexive 'Zibun'

Compared ta English reflexi ves such as himself or herself,

the Japanese reflexi ve has unique properties.

overview of these properties (Fukui, 1984).

summarize thern below.

Fukui gi ves an

1 will briefly

First, zibun is used regardless of persan, gender, and number.

Secondly, the antecedent of zi bun must be a human bei ng. The third

-32-

1

T .

charactcristic is that /ibun must ln' hound by .t \1' \\hjeh IS fi

subject of the clause. (nli~ \\ill lH' dl~vd(}pt'd in St'ltion 1.:ï.)

Moreover, tht~ tapie phrasl~ cll1d 1 h\' ht'dd of .J n'Idt Î\'(' (1.1I1!-.(~ (\10

also become an antcccdt'Ill of /ibull. l list ly, i t (',Hl tw disnmn,(~

baund; i.e., Libun need Ilol have its itnlf'Ct~dcnt 111 IlH' st'ntenn!.

since i t ran get i ts rcfcn~nce from an NP whi ch clPpt'drs in sorne

preceding sentence in the discaurst.'.

Expanding the above propt~rtics given by Fukui, 1 need ta

mention tv.o more ff:atures v.hich an~ very import.ant in the present

experimenl. Since null sllbj(~ct.s are often allo\\(~d in .Japdnc!-.(!, the

speaker can be the an tf!ccd(~n t of /i hun. 1 n f <le t, t hl! fi rs 1 pt?rson

subject (\\atashi '1'. ctc.) is USlIci))Y omil.t.cd. ~lJrtlH'rmon!, even

in the sentence where t.he subj(~ct is not 'l', t.he s(Jedk{~r can becomc

the anteccdent of zibun. This may be rdated la UH~ faet mt~nt.ioned

by Fukui that the t.opic phrase can be th(~ anl(~ccdcnt. Ihis learl~ to

one more feature, which is that zi hun ('an even h(~ ('urd (!rcnt ial wi lh

an NP whirh is a discourse topie but docs nut. oVI~rtly appear in the

discourse.

It has been argued that various factors may affect the

interpretation of zibun ('internaI feeling' and 'awarcncss' (Kuno,

1973a,b), 'point of view' (Kuno, 1978), clc.); howevcr, 1 would like

to foeus on the grammatical constraints on the binding of /ibun In

the follawing section.

1

1

t

3.4.2. Governing Category for Z i bun

Among the five classifications suggested by Wexler and Manzini.

Japanese belongs to the (e) type; therefore, its reflexive zibun may

get i ts reference from more than one ~p in (18) 1 •

(18) a. [John-ga, zibun-o, seme-ta.]

John-nom self-ace blame-past

'John blamed SELF:

b. [John-ga, [Bill-ga 1 zibun-o'/l shokaishi-taJ to i t-ta,])

John-nom Bill-nom self-ace introduce-past comp say-past

'John said that Bi 11 introduced SELF:

c. [John-ga, [zibun-ga, Bill-o l shokaishi -ta] to i t-ta.

John-nom self -nom Bill-ace introduce-past comp say-past

'John said that SELF introduced Bill:

d. John-ga, [[Bill-ni J zibun-o'/l shokaisuru]yoni] tanon-da.

John-nom Bill--dat self-ace introduce comp tell-pasto

'John asked Bill to introduce SELF:

Since (18a) is uniclausal, zi bun is bound by John. When a sentence

is biclausal as in (18b), zibun becomes ambiguous since it is either

bound by Bill (which is a clause-mate of zibun) or John (which is

not in the same clause as zibun). This sentence shows that zihun

can he long-distance bound, as predicted by value (e) of the

parameter. (18c) shows that zibun can also appear in the subject

-34-

1 position and it C;Jn be büund 10 .ln \1' \\hich is outsidt' ut ils

clause. (13d) is alsa dmbiguoll!:-' in th,iI /ibull (';\11 rd.'!" l'ilh.'r 10

Bill or to .John. Although Hill is .Hl indin'rI ()h,it'('\ \\hit Il is

marked dati ve case by ni. 1 assumt' t hl! s lruct urt' of (1 Rd) is

similar tü an English equivalent (~.f.. (lk)) in th(lt Bill controls

PRO in the embedded subJcct position. lh(~ following lre(!::-; ::-;how the

structures of (l8b) and (18d) 2.

IP

NP~ l' c:::\ /'-..

John,-ga VP 1 /'\

CP V -ta Ipast'

\C' it-

Ip/\C

NP~ J' ta

Bil~ VP~I Np./"'\ V -ta · past' ~

zibun./ )-0 shokaishi-

IP Î"--.

NP J' c3 ~

John, -ga VI> 1 ~

NP CP V -da 'past'

B~ll . \,. 1 J nt C tanon-

IP~C /\

NP l ' yoni ~ /\

PRO) Vi> 1 /\

NP V ~

zibun,/)-o shokaisuru

Both sentences are made up of two clauses; hence. Li bun lakes ci ther

a non-local NP (main subjcct) or a local NP (embcdded subject) as

i ts antecedent.

1

1

3.5. The Proper Antecedent Para me ter

There is another parame ter that "e need to consider. It is the

Proper Antecedent Parameter which is suggested also by Wexler and

Manzini (1987), It has two values wÎ th respect to what is allowed

as the antecedent of the reflexi ve in the language, i.e., subjects,

or subjects and objects. It is defined as (19):

(19) Proper Antecedent Parameter: A proper antecedent for a is

a. a subject

b. an element S whatsoever

(Wexler and Manzini 1987:64)

As 1 have mentioned in Section 3.4.1.. zibun takes only subject NPs

as i ts antecedent.

(20) a. John-ga Bi Il-ni zi bun-ni tsui te hanashi -ta.

John-nom Bill-dat sel f -about

'John talked to Bill about SELF:

talk-past

b. John-ga Bill-ni zibun-no syasin-o mise-ta.

John-nom Bill-dat self -gen picture-acc show-past

'John showed Bill a picture of SELF:

The above sentences are both uniclausa 1; thus, John is a subject NP

white Bi 11 is an object NP. Since zi bun can be coreferent only with

-36-

1

,

subject ~Ps. the dnteCt'th'nt uf /ibun in C20,d ,md (.20h) I~, .John bllt

not Bi Il. T1}('n~fon~ .. ]d(),!npSt l hdld\'t)~ 1 i kt' .1 t "pt' ('Il 1.111~IJ.I~(\ in

(19) .

Engl i sh, on the otlier hand. bl' longs t 0 t Y[ll' (h) LlIl~l1"Rt's. The

reflexive in the sentenCt~s bclo'l\ is ,lmbiguolIS SI net' il may rf!fcr t.o

either a subject NP or .ln object. ~P.

(21) a. John talked to Bill about himself.

b. John sho\\ed Bi Il a pictun~ of hi msd f.

Thus, with resp(~ct to t.his paramd.f!r. ~.ns~lish has more

possible antecedents lhan .Japanese ducs.

Wexler and Manzini (11J8ï) lhus propos!! I.hal Binding Ihenry

contains not only onf~ bul Iwo diffcrenl pilram(~krs: Ihe (ioverning

Category Parame ter and t.hf! Propf:r Anit('f~(iPnl P<lrdrn(!h:r. They daim

that these t wo parame t ers ex i::... 1 i mkpcn(kn t.l y (cl n: 1 a t ion (kscr i bed

as the Independencc Principle) althollgh Ihey inh!ract \\Îlh respect.

ta the behavior of rcflexivcs. \\ill considcr Ihis proposai in

greater dcpth in the next chapter.

In this chapter. t'l\O paramelf!rs have been invcstigated:

the Governing Calegory Pararnelf:r, rlnd thl: PrOpf!r An t.f!n!df!nt

Parame ter. The former is five valw!d pararnctf!r and Ihe latl.f~r is

two-valUf~d. Al t.hollgh UH'SI> (J,lrilml:h:r:> arc propo~('d to f'xplain bOUl

refl(~xivcs and pronomindl!-., 1 hdVf! f(J(,llscd on rd !Pxives. Principh!

:~ 7

1 A of Binding Theory is also reviewed. It has been shawn that

English and Japanese are at opposi te ends of the scale in their

values for these two parameters. Japanese has more possible

antecedents for the reflexi ve on its setting of the Governing

Category Parameter, while the situation is reversed in the Proper

Antecedent Parameter: i.e., Eng 1 i sh has more possible antecedents

for the reflexive.

{ -~

~ -:~ - ;:-~- ~~-_',.... __ ---::;;r~ ~~_:;--"-: ~.. ::_r ~,..~ ~-

.::-.., ::. ~

...,.!'-;::::--=- --

,~-

" -;:: -, '

-~---;;,- , ~

~ ~-~~ 3

!r.-; t_ --

c

-38-

1

1

Faatnatcs ta Charter 3:

1. The following abbrcvintians arc Ilsed in Japancsl' ("<impies: nom

nominative, ace: ::: accusative, ciat " dative, gl'n - gcnitivc, ncg - negativ<'

comp :;: comp)('mcntizer.

2 . 1 t i s still a d c b a LI bic i s S Il (' w h c t Il c r 0 r no t J a p il n f' s cha san 1 N H. n u d ('

(c.g., Fukui (1c)8(') daims that thcrc: an~ no fUnl'tion,d c<.Ilcgoric::o. in

Japanese; thus, no INFl; Kitagawa (198(,) daims that tlwrc is IN~L.) Sinet'

therc i s no subjcc t - verb agrc('m en tin .).l!ldIlCSl', no AfiH l',ln hl: t1ssumt'd.

However, 1 a~sume hcrc thdt HNSI· I~ Ml IHI ,1lthOIl~h Il',IVi' III<' i SSlIl' of

whcther or not thcr(' i!'-. an infinlllvdi open, dS this i:-.su(' will Ilot bl'

crucial for the stuciy ln titi' thcsis (JIll' puint thdt 1 would lib' 10 nutt'

here is that therc IS a difff'r('nc(' b('\\',('('n III\' c'mbcddt'd v('rb~ in (lab ) and

(18d ): shokalshi-ta 'intrudurc pM,l' in (lSb ) is inflcch'd cl~ pd:..1 whilc

shokalsuru 'introdur.e' in (lad) IS not Inflerlcd. This "ICk of infl('C't ion

may be similar ta I:ngll~h inflnitiv,l! l'on· ... trllction~. ,tllhough Ihr.rc i::o. no

overt clement Ilke ~_ngllsh 'to' to Inirodill'I' infiniliv,lI Ct.lUSCS in

Japancsc. 1 assume hcr(' lhtlt yunl 1:'" d cOIDPll'm('nll/cr, liS 111C'rc' are cases

wherc inflcrtcd vcrbs attdlb tu -yoni l'fll' eXdmplc,

I.John-gaIIBill gaYdt lalY(Jnllvat tdlll

.John nom Bill-nom do-pa:...t

'John dld (it) a~ Bill did (it)'

ln arder tü dctcrmine thcsc issues, furthcr rcsearch is ncedcd.

1

1

Chapter 4

LEARNABILITY AND THE SUBSET PRINCIPLE

In this chapter, l will consider hOIN linguistic theories can be

linked to learnability of a language ln LI and L2 acquisi tion. The

'Subset Principle' which \\as originally proposed by Berwick (1985)

and by Wexler and Manzini (1987), \1anzini and Wexler (1987) will be

introduced. The Subset Principle has been proposed as one of the

learning principles in LI acquisition \\hich is independent of the

principles and parameters of UG but interacts wi th them. It

provides a means of resol ving the problem of lack of negative

evidence in Ll acquisi tion: i t predi cts the order of the chIld's

hypotheses about the target language when parameters involve values

which stand in a markedness relationship to each other.

Assuming that the Subset Principle acts in LI acquisi tion, we

may then ask whether or not it operates in L2 acquisition. The

experimental studies in the literature suggest that the answer 1S

negative. 1 will consider how the Subset Principle relates to L2

acquisi tion of refl exi ves \\ i th respect ta the Governing Ca tegory

Parameter (Gep) and the Proper Antec~dcnt Parame ter (PAP) discussed

40-

J

,

in Chapter 3. Previous st.udies relatcd to this que:;.;t ion wi Il also

he examined.

4.1. The Subset Principle

In Chapter 2, 1 have shown that lack of n~Rativc evidcnrc is

problematic in that there is nothing t.hat oblig(~:;.; IIH' ('hild to

retreat when he or she hypothesizes an overgenerdli7<,d grilrnmar. For

exampl e, when lhere is a parallIeter wi th fi ve val Il!!!;, <1::-' i Il 1 lit! GCP,

i t may be asked how a chi 1 d whose t.arget 1 angugc J::-' él tyP(~ (il)

arrives at the right grammar wilhout mistakenly hypoth(~!)i/jm~ !.hal

the target 1 anguage i s one of lypes (b), (('), (d) (mcl (').

The Subset Principle has beerl proposed ln solw' t.his kind of

problem (Berwick, 1985: Wexler and Man7ini, 19R7; Man/Îni and

Wexoler, 1987), Berweick (1985) noles:

ln the special case where one tanv~t langlJas~(~ is proI)(~rly

contained within another, lhe point. of t.itis (,(HldiLion is to

ensure that the acquisition pro((!dlln' éJIVlrlY~; S~tIf!:;S(~S (j

subset 1 anguage if poss i hl (!, t.hé! t i!, Il H' : ,/II ,il 1 (!!, t 1 angllag(!

that is also compalible with t.he ,,(J~;JI ivr~ ('vidcm{' so far

encountered. (1985:236)

The Subsct. Principle states that. t.b(~ (hild c:h()()s(!s lhe p,rammar

p,cncralinp, th(~ sméJIJc:d. l;Hlp'UiJp'(~ fin;!, and pr(J('(~(!(h 1.0 p,o hf'youd

- 4 l

1

1

1

that val ue only when posi ti ve evidence of the larger language is

available.

Following Berwick, Wexler and Manzini c1aim that the Subset

Principle plays its role only when the Subset Condition holds. They

formulate i t as follows.

(22) Subset Condition: For every parameter p and every two values

i, j, of p, the languages generated under the two values of the

parameter are one a subset of the other, that is, L( p( i )) ç L( p( j))

or L(p(j)) ç L(p(i)) (Wexler and Manzini, 1987:60),

The GCP and PAP meet the Subset Condi tion. The values for

these two parameter setting show an · entailment' relation as

illustrated in (23).

-42-

1

1

, J

(23) a. Governing Category Parameter - .,~.- -

Parilmeter

---- ~ _ r _ 1_- ~ - L-_

t

1

The grammar represented by the smallest circle is contained in the

grammar represented by the larger circle (Language(a) ç L(b) ç L(c)

ç Ud) ç; L(e}). Thus, the data which rnotivates the smallest grammar

is also compatible with any other grammar of the larger circ1es.

The Subset Princi pie i nvol ves the notion of markedness.

According to Wexler and Manzini, the parame ter setting generating

the subset is unmarked and the setting generating the superset is

marked. Hence. the Subset Principle predicts that the child

hypothesizes the most unmarked grammar first before hypothesizing a

more marked one. They suggest that the Subset Principle is part of

a learning function \\hich is not contained ln UG. but that it

interacts \\Ïth the principles and parameters of UG. Furthermore,

markedness hierarchies of the values of parameters are to be

calculated by the Subset Principle. making it unnecessary to state

these hierarchies wi thin UG. Thus. as a learnability function, the

Subset Principle guides a child to compute the markedness hierarchy

and ta select the parameter value leading to the most unmarked

grammar as an initial assumption.

Concentrating on acquisi tian of anaphors, the Subset Principle

will predict the following acquisition orders. For English speaking

children and Japanese speaking chi Idren acquiring reflexi ves. the

mast unmarked value of the GCP, value (a), will first be assumed.

English learners will stay there because there will be no evidence

that indicates the possibilities allO\\pd by the grammar (b)~(e).

Japanese learners. on the other hand. proceed to hypothesize value

-t14-

1

1

J

(e), as there should be plenty of posi 1 ÎVt' t'vi{knn~ in Il)1' input for

this value. The situation \\ill hl' rt'\'(~n·,\'d \\h\'11 lllt'y ('ompllft~ IIH~

value of the PAP. Again both ll'.lrnt'rs sl.lrl l'rom v,lllJt, Cd.

assuming lhat the anh'!ccdent. of the rcfkxivc should ht~ il subjcc!,;

only English learners go on la hypoth('si/l~ thal an ohject can <llso

be the antecedent, corresponding lo valUl~ (b), the wider grammar.

We should also note herc !'hat. one language is nol. necessarily

the smallest. language or lhe Iclrg(;st language in both respect.s.

English chooses the smallesl value for the GCI>. but dHJOS(!S the

largest. value for the PAP; .Japanesc IS complel.ely opposite: it hrlS

the largest value for th(~ f1CP and 1 Ill: smallcsl vallH~ for t.he PAi>.

Wexler and Manzi ni propose t ha t ' 1 he subsct n:l a tions bpI, we(:n

languages generated under diff(!n:nt vailles of " pdrarneter rcmain

constant \\hatever thl; values of 1 ht: 01 !J(~r pdramd.t;rs <Ire 1 dk(~n 10

be' and name this propprt.y th!; . Indl'flt~ndcncc Principlc: (W(;xkr and

Manzini. 1987:65)

do not interfere

independently.

Thus, the t-wo subsct condi lions consÎ(h!rcd herc

wi th each otht:r; HiC two parameters hold

ln sum, in LI acquisi tian. the Subset Princi pie prcvcnts the

child from hypothesizing the wrong grdmmar; in consequence. his or

her grammar is fn!c from (~rrors ('al1~(!d by ovcrg(!ncrdli/atiun. As

far as rcflcxiVf~s arc concerrH!d . .Jakubowic'/ (19R~) 1 ObS(!rVf!S thal

children whose targd languélgf! i~ ~.ngli~h corn'cUy hind rdlexivf~s

to the local antf'('('dcnt (aIs!) Clllf'll dnd Wp'df'r, JCJR7; [)f!ut!-iC'h,

Koster and Kostr!r, )(J8f); Wcxkr and Chipn. ]fJRS}.

1 :l

1

f

1

4.2. The Subset Principle and L~ Acquisition

What can the Subset Principle say about L2 acquisition? Is it

acti ve in L2 acquisi tion as i t is in LI acquisi tion? Or is i t

inactive? Several studies have been conducted to examine whether or

oot the Subset Principle operates in L2 acquisition. And indeed,

they suggest that the Subset Principle does not lead the learner to

assume the most restricti ve grammar from the beginning (Whi te, in

press; Zab!, 1988).

4.2.1. Studies by White (in press) and Zobl (1988)

1 will consider two studies \'vhich investigate the situation

where the subjects' LI represents the superset value and their L2

the subsct, as below.

(24)

-46-

1

,.

This situation is crucidl for hoth !'.tlldit'S !'.WCI' IllI'v ,1r~llt' Inr tht~

LI transft;r h~polht~sis: i.t'. instt'dd Id ,Hqllirinp; tilt' slIh!'.t'l grtunlll,1r

straightfon\c1rdly dS 1 lit' Sllb~t,t Prin(ïpit' IWI'dicts. 12 !l'drrwrs

transfer their Li paralDt~ter vall1l~ cmd [,Iii tn arriv<' clt tht' 1.2

grammar.

White (in press) investi~at~s thl' adj,iCcncy (ondition on Case

assignment; t.nglish rcqu1n~s i tstrkl ad,iaccncyl so t.hat no

adverbial may inlerVf~ne hchH~en t.he verb ,md 1 he object. w\d le

French is [-strict ddj.-H'(~n('y J so thal somt~ adverbials (i.t~. ID,mner

adverbials) can be placcd t)(~l\\{~en tlV' 1\\0. Whitc's sub.wcts an~

both French learncrs of t.nglish cmd l'nglish learners of I·n~nch.

Resu lts f rom var IOUS tt~S ls. i.e. gramma 1 i ca 1 i t.y .i\Jd~cm(~n t. lasks

(paced. unpaccd. preference). show 1 hcl t many ,. n~nch h~.-lrn('rs trea l

English as if il ~ere fn!nch, viol,Itinp, [Ist.riel d(Uc)('(~n('yl. The

English learners of f'rcnch seem 10 undersLmd Ihe [st.rid.

adjacency] condition of french parli,illy bul not compkldy. From

the behavior of lhese groups, Whi le fa vors the t ransf('r hypot hesis:

L2 learners transfer their LI valtlf: in t.he i}('quisit ion of 12. which

leads to fai 1 ure to arri ve at the corn~ct tdr~et. grammar.

Zobl (1988) examines the [+configurat.ionalityl paramcter using

Japanese 1 earners of Eng 1 i sh. II(~ assumes tha t. the suhjec ts' I.l ,

Japanese. lS l-configurational J and t.heir 1.2 [Ie'onfigurationall:

since [-configurationall is UH! supp.r!'.ct, t.he .Japa!1(~Sf~ value is Ul(!

superset of lhe r.nglish on('. lIis tl!'.~llmpt ion IS t.hal a

[-configurationall language lacks il \'l' nf)d/~ rll1d hiJ~ rl fIat. struct.un!

·t 7

t

l

1

yielding very free word order and that, in contrast, a

[tconfigurationalj language has a VP node and a hierarchical

structure. so that word order is relati vely fixed. Subjects were

asked to place adverbials in whatever position they fel t the most

sui table for each test sentence. Thus the test si tuation resembles

that of Whi te. Zobl found that intermediate level learners failed

to observe English as [+configurational J while advanced learners

were more accurate in observing i ts configurationali ty. He daims

that, although the Subset Principle does not operate. learners are

subsequently able to reset to the English value correctly. 1 will

not go into his analysis of the causes of resetting, but simply

mention that his study sho"s the initial non-operation of the Subset

Principle.

4.2.2. The Study by Finer and Broselow (1986)

There is one small pilot study which investigates the GCP in

the acquisi tion of Engi ish reflexi ves by six adult Korean subjects.

Korean is like Japanese in that the reflexive can be bound in the

whole sentence; hence, i t is a type (e) language for the GCP. The

reflexi ve is only bound by the subject; thus, Korean is a type (a)

language for the PAP.

At the time of testing, the subjects were students in an

intensi ve Engl ish program at uni versi ties in the Uni ted States. A

picture identification task was conducted in which subjects were

- 4 8 --

shown pairs of pictures. Tht' subjl'l h Iht'n ht'.lrd d !-,l'lllt'll( l' .1Ild

t were askcd to indic(Üt~ \\hich of tht' 1\\0 picturt's \\.IS ,lpPFopri.ltt' for

the sentenn'. or both pidurt's if hoth \\i'n~ flppropriatl'. 11lt' ft'st

sentences wcre of the follo\\ing t\'.o tyP(~S; edch type was n'presenlt~d

by f our sen tences. 2

(25) a. Mr. Fat thi nks that Mr. Thi n \\ i Il pai nt himself.

b. Mr. Fat wants Mr. Thin to pain!. himsdf.

The resu 1 ts show lha l Korean \t'(lrnc~rs assumed U}(~ 1 Dca 1

antecedcn tin the t enscd cl auscs bu t lIf t en rd i led t 0 do so in lhe

tenseless clauses. Overa II responst'!-, fllld 1 he i r pcrClm t "ges ,1rn

shown in the next table (h n(~r and Bros(!low, 198b: App(~ndi x B).

(26)

~ocal Non-local El Uler -~- ..... ---- ~ ..... -,.,....- --~-~

(Tensed Clause]

?? ') 0 ... "'" ...

91.7% 8.31 0% ---- ----- ---

[Infini tlve Clausel

14 9 1

58.3% ,7.'5% 4.21 ~-- -~~--- --- ----- --~--

Total ,b 11 1

t

l

f

Finer and Broselow interpret the n~sult as indicating that the

learners have picked neither their LI value or L2 value, but an

intermediate value of the Gep, because it seerns that the learners

distinguished [±INFLl as taking the local antecedent in the tensed

clause but rejecting it in the infinitival.

However, they also offer an alternative interpretation ln which

the Subset Principle do es operate perfectly. Accordi ng to this

interpretation, the learners assurned the rnost unmarked value for the

Gep and for the PAP; ihis led to a cornbination of both settings.

That is, ln a tensed clause structure like (25a), the subjects

correctly chose the local antecedent (91.7% of the responses); on

the other hand, ln an infinitive clause structure like (25b) , they

analyzed Mr. Thin as the object of the sentence and avoided it,

incorrectly choosing Mr. Fat as the antecedent (58.3% of the

responses). In each case, they chose the rnost unrnarked value for

the parameter. If this is really what the learners are doing, Finer

and Broselow claim that the Subset Principle may be fully activated

for L2.

The second interpretation cannat be confirmed on the basis of

their resul ts since they did not have sentences designed solely ta

test learner's setting for the proper antecedent. For such an

analysis, we need to check if the learner chooses only the subject

and not the object in a simple clause sentence.

1 am also skeptical about their fi rst i nterpretation. That is,

al though they argue that the subjects picked an intermediate value

-- 5 0 --

1

1

•

for the GCP, Mr. Fat and Mr. Thin an' hoth conn'jvahl(' anh'('(·(knl!-. ---- -----

for himself in ihe sentencc. If lhere is ~Olll(' strong tt'n<i<'ncy in

the subjecls' LI ta pre fer Mr~ J'il_l in a scnknn' lik(' (2Sb) rdther·

than one like C~Sa). ihis rcsult rnay ht~ tr.l('c.lhl{· lu 1 J. (1 will

develop ihis idea of 'preference' in Sectioll ,1.i.) 1 n ()rd(~r t ()

determine whcther or not t.h('y pick(~d t.he int(!rmcdicll<' V.t1IH!. WC! ll{,f!d

ta test more complex structures sllch as the followinP,":

(27) [John says that [Mr. Fat asked Mr. Thin [PI:D lo paint himself.1 J 1

If L2 learners pick an interrnediale value, on the hasis of whether il

clause was lem:ed or not. tlley should not d\Oose .John ilS t tH'

antecedent of himself in (27). If il turns out thdl UI('Y allow the

non-local antecedent. John, as lhe anl(!('t!dcnt, wc must concJudc t.h(tt

they are not choosing the inlermedi(Jtc, hui I.lte larp,c:·;t valll(' of the

Gep, as in lhei r L 1.

4.2.3. The Study by Thomas (1988)

Thomas (1988) also inveslip,ates the riCi> and Ulf~ PAP wi t h

learners of Engl i sh. The subjeC't~) wcn: nd 1 j V(' spcakf:rs of 20

diffenmt languagf!s. She found !hat 1('(JrIJ('r~) of 1·.np,li~~b did not

these

:; 1 -

antecedents; this preference was also observed in nati ve speakers of

English.

Thomas's test sentences consist of four types: two for

examlning the local antecedent. and two for examining the proper

antecedent (or what she calls · the subject strategy'; she assumed

preference for the subject antecedent among nati ve speakers of

English. as suggested by the previous literature). AlI sentences

contained tensed embedded clauses (i.e. no infini ti ve clauses were

involved). Sorne of the test sentences were designed to see whether

the learner chooses the local subject antecedent even in the

si tuation where the non-Iocal/object antecedent reading is more

plausible pragmatically. Examples are gi ven in (28),

(28) Type 1 Ernie was sorry Cookie Monster hurt himself.

Type 2 Mary angrily told me that Sue had spilled a lot of

paint on herself.

Type 3 Susan gave Mary three photographs of herself taken

Type 4

last summer.

After the medical tests were completed, the doctor

informed Bill about himself.

Type 2 and Type 4 are the ones designed to mislead the learner.

That is, Type 2 can be used to show whether those subjects who

correctly bind reflexives ta local antecedents will violate this

local antecedent condition under pragmatic pressure. Type 4 can

-52-

1

,

•

antecedents in Type 3 \\ill stick ln Ihis sllh.it~ct condition \\ht'n non

subject ~Ps an~ pragma t ir,d 1 y Ll\;ort'd.

sum up Thomas's rrsults in Table (2!JL (They an~ originally

given in her table 1,,1.)

(29) L2 L~?rne!,"s Control Gr_<?!lP -~ --~-- ._-

local non-local f~ i ther local non local either --- --------

Type 1 65.49 Il. !Ji 22.20 99.27 .b4 .18

Type 2 48.60 r -9 ').1 1 S.bI !JI. if> 8.0C] 0.55

subject abject ci ther ~l!bJ~~_~ object ci thcr ----- --- ---- ------ -

Type 3 55.92 9.47 i4.hi -? ~p I ...... l ...... 2.4") 24.73

Type 4 22.55 48.64 28.92 19.73 154.18 26.18

(Figures are rnean pcrœnts. )

Overall results show that the di ffcrences bet.ween L2 learners

and the control group ln Typ0, and fype 2 responses are

significant, which suggests that rnost. lt;arners set. the value for the

governing category wider than it is supposed t.o bf!. As far as Type

1 and Type 2 an~ concerned, the Subs(!l Principl(~ dOf!s not secm 1.0

operate. Furthermorc, these rcsults an~ inconsisl.Œt with Hner and

8roselow's proposai t.hat. 1.2 learners adopt an inl.ermf!diale value of

the Gep, sinC(~ aIl test scntpn('(~s WCt'"f! tcnsed. Thf! int.f!rmcdial.f~

values of t.he GCP requin! local clnlf'('(!(knts in Icnsf~d scntence~ .

1

1

1

1

The results from Type 3 and Type 4 suggest that L2 learners and

native speakers respond in identical ways; Le. both p;roups bind the

reflexi ve ta the subject in neutral Type 3, but bind i t ta the

object in Type 4, which is affected by pragmatic factors.

Since Thomas~ subjects consist of 92 language learners of 20

different LI backgrounds, i t is not c1ear whether they transferred

their LI or not. However, she gi ves an interesting comparison

between Spanish learners and Chinese learners, the two largest

groups with the same LI. With respect ta the Gep, Spanish is like

English, while Chinese is like Japanese. Thus, even if the Subset

Principle does not operate, but these learners transfer their LI

instead, Spanish learners should be able to find the local

antecedent which is the correct analysis for English, while Chinese

should, wrongly, find the non-local antecedent. Thomas suggests

that this is not the case, as is indieated by the following

compar i son.

( 30)

Type 1

Type 2

Spanish LI

local non-local

59.48 18.90

49.90 37.55

either

2153

12.55

local

69.04

49.04

Chinese LI

non-local ei ther

7.29 23.04

34.29 16.67

(Figures are mean percents.)

None of these differences between Spanish and Chinese groups were

reported as significant. Therefore, the results from the Spanish

-54-

1

•

transferred their l.i ~rall1lJl.lr.

4.3. Preference vs Grammdr

ln Thomas's (P)8~) slll<l} , the Ildlive :.-;peakers of l',nglish tt'ndl'd

to choose a subject ~p <lS an ant(,l'('dl~nt of tht~ rdlpxivc ratller théln

an abject ~p (over 72% of 1 IH~ t i me') in .1 s(~n lt'nce 1 i k(~ (il),

(31) Susan gave ~1(lrv thn'(' phot()~~raph~ of h('rself takcn Llsl summer,

Althaugh r.ngl i sh a Il ows oolll SlIsan <1I1d \-1dry as possi bh! ant c('(!(knl s

for herself, t.hcy \\f~rC nol ('hOSt'l1 pqudlly, lholll,'s int(!rprds this

resu 1 t as bt~ i ng duc lo the f é)(' 1 Ulil t. èVt~n wh(m t.h(~ Rr am mclr il Il ows

two possibililics, il is likcly Ihdl sp(~akc:rs of Ih(\ langlldge have

sorne sent.f!nce inU:rpn~tali()n slral.t!g) \\hich LJvors one of th/: t,wo.

would like lo n~ff~r 10 Ihi!-. pl)f!nOrnfmUn c1!-. ·pn:fenm('{!'.

to favor one possibility among sf;vpral i.hal are predict(~d by d

grammatical Ulf;ory, Therefon:, Ihe f d( Ilhat possibili t i(;s iln! not.

pref errcd by spcakf;rs of 1 hr\ 1 dnp,lIdRI! dof'S nol Ill(!an t.hcy tire bad in

the language. Il is ~imply Ihe ('(i!',(! Ihd! Ulf: Ul(!Ory itsl'lf dom; not

predict which possibility Illay t>f' f(}von~d OVf!r UJ{~ othf~r(S). Thp, PAP

states tha! for tin rny,li~h !-.('n!e!1( " likp (11), two NPs, SUSrHl ,md

.) ;)

1

1

1

study shows, nati ve speakers chose Susan ta be the antecedent more

often than Mary in a pragmati calI y neutral sentence. That is,

native speakers of English had a preference for subject antecedents

over abject antecedents of the reflexi ve. Only 24.73% of these

subjects abserved the sentence ta be ambiguous, a result which

cannat be predi cted by the PAP.

Another study (Nari sawa, 1988) also daims that nati ve English

speakers have a preference for subject antecedents aver abject

antecedents ln sorne cases; mareaver, L2 learners, especially

beginners, may transfer this LI preference to L2. Narisawa examined

the PAP using native speakers of English learning Japanese. In this

case, the subjects' LI is marked and the L2 is unmarked. She also

had a control group of native speakers of English. Her subjects

were considerably accurate in finding subject antecedents for zibun

in one type of test sentences (e.s. Debusan-wa Yasesan-ni zibun-no

koto-o hahashimashita. 'Mr. Fat talked to Mr. Thin about SELF.'); in

contrast. they were Jess accurate in the other type (e.s. John-wa

Bill-ni zibun-no uchi-de aimashita. 'John met Bill at SELF's house.')

Since the control group showed a preference for subjects ln the

former type, but indicated the sentence to be ambiguous ln the

latter type, she states that the learners seemed to simply transfer

their LI preference.

As far as the Gep IS concerned. native speakers of Japanese in

a small study (a preli mi nary study for the experiment in this

thesis) favored non-local antecedents over local antecedents in

-56-

1

,

general. in cases \\ht'rt' tilt' gr.lmm.lr sLltt's noth .In' possihll~

antecedents. T1H~ s{~ntt:nCf~ hl'In\\' ('ontains /iblill in tht~ ('mbt~dd(xi

object posi tian.

(32) Watashi~wa[~Yukiko ga Libun no adclOd a kinittdrulto omottdla.

l-topic Yllkiko nom self ?;('n nickname dCC like ('omp lhollghl

"l thought that Yukiko likcd Sf·.IYs nickname.'

11 out of 13 subjects chose wata~bi 'l', a non Iorai sub.i(~ct., as the

antecedent for 7iblln, one cho~(~ Yllkiko, a local subject. and one

responded that lhe sentence was amhip,lIoliS. lIowever, for two out of

the seven sentences used in t.he s llldv, 1 hcrc was an i ncrCdSC in t he!

number of responses favoring Ihe local antf~('{~d(ml or ambiguity.

This suggests lhat a prderf'ncc for cl p,iven antf~cc(knt IS infltJ(~nced

pragmatically, most likely, by lhe mCilning of li If! verb. Ihe last.

two cases \\erc t.he scntences wi t.h the (:mbcdded vcrb, syokaisuru

"introduce' and wakaru ·undcrsland'. (fhe ot.lIer vcrbs IJsed in t.he

study wen~: t}o~orini~~omou 'bf! prolld o{', propose surll ,

propose,

naguru 'hit'.) The experiment.al study presenled in t.his thesis is

designed ta examine if t.here an: v(~rbs which makc the speakers

observe a preference for non ~ 1 oea 1 an tcccden t.s or 1 oca 1 an tcn~dcnts,

ar bath.

57 -

1

1

4.4. Hypotheses for the Experimental Study

Summarizing the discussions in this chapter. 1 would like to

list several hypotheses with regard to my experimental study on the

acquisition of English reflexi ves by native speakers of Japanese.

The fi rs t concerns the Gep:

Hypothesis 1: The Subset Principle operates identically as in LI.

This hypothesis predicts that Japanese learners

successfully arri ve at the English value. as i t is

the most unmarked.

Hypothesis II: The Subset Principle does not opera te properly.

Japanese learners transfer their LI value yielding

the wrong setting for the L2 grammar. This predicts

that Japanese learners bind the reflexive ta the NP

which is not allowed by the Engl ish grammar and fail

to pick the correct English value.

Hypothesis III: The Subset Principle does not operate and LI

transfer does not occur ei ther. This predicts that

learners choose neither value (a) nor value (e).

Their behavior is distinct from the LI and the L2.

They somehow pick a value in between.

-58-

1

J

T

Hypothesis IV: Learners t.ransft'r tllCir pn~fen~rlCl~ in lIt 0 IllI' L2

grammar. This mean!,-; t helt. even \\II/'n the p,rammar

yields t.he largest \dlue so that ,II1Y NP in the

sentence can be t.he anteredent.. if there is

preference for some NP among thcm. l.he 1 earners

transfer that preference. Thus. errors can be

predicted by LI preference.

Hypothesis V: UG is not accessi ble to L2 learners and learners

will produce an unnatural grammar whirh is not

allowed or explained by the principles or parameters

of UG. This firth hypolhesis t.akes the most radical

posi tion against the operation of UG. 1 t predicts

that learners arrive at an unnatural grammar, such

as one t.hat will admit only non-local antecedents

but not any others.

Among these, 1 wi Il adopt Hypothese~_ll_~_~~ _J~. hypothesize

that Japanese learners wi 11 not i mmediatel y arrive at the correct

English grammar. Their ini tial setting will be affected by their LI

value. Moreover. if there is any preference for one possible

antecedent over others. they may transfer that preference.

Hypothesis III is the one that prcdicts Finer and Broselow's

results. As proposed ln section 4.2.2.. an examinat ion of how

native speakers of .Japanese behave \\hen judging .Japanese may show

-59-

,

1

t

1

that such native speakers treat the reflexives differently in two

diffèrent si tuations, l.e. tensed clauses and non-tenscd clauses

(equivalent to English infinitivals). If so, it may be concLuded

that a preference in the LI is transferred. If this kind of

preference is revealed, it indicates that Hypothesis IV may override

Hypothesis II.

Evidence for any Hypothesis from 1-111 will be evidence against

Hypothesis V, since Hypotheses I~III are based on the assumption

that UG is available.

As for the PAP, Hypotheses 1 and Il may predict the same

results. As Japanese is the smallest grammar, bath hypotheses

predict that Japanese learners start from value (a) and. if there is

positive evidence. will reset to value (b). Hypothesis III ~il1 not

be applicable here since there are only two val ues (no i ntermediate

value). As there is only one value in Ll (i.e. preference equals

the grammar), hypothesis IV will predict that learners pick value

(a), ju,;t as Hypotheses and Il do. What may be different is that

they stick to value (a) even when there is positive evidence to the

contrary, if value (a) IS their preference. Hypothesis V predicts

that learners choose an unnatural language; perhaps taking the

abject as an antecedent but not the subject.

Experimental results for this PAP setting may reveal whether or

oot learners avoid object antecedents because their initial setting

for the PAP is the unmarked one. If they actually avoid them, theo

Finer and 8roselow's interpretation (Le. the learners pick the

-60-

smallest grammar for both the GCP and the PAP) May prove to be

t

J

1

1

,

feotll-otes ta Chapter -t:

1. Jakubowicz aisa examines acquisitiot\ of prot\ot.tns and claims that younl

c:Mldrcn trcat pronouns as rcflcxi'vcs: howcvcr. her analysis in which

reftexives and pronouns are trcated as one catclon does not seem so

promising.

Z. finer and Brosclow also cxamincd sentences with pronouns; howcvcr,

wll1 not discuss those resul ts hcrc.

1. Sentences such as the following. as well as thosc with control verbs.

"K and tell. arc includcd ln the cxpcrimcntal study.

(a) \4r. Fat y,ants l\1r. Tt)}n. ta palnt him:o.clf ,II .

.!.!nl is an EC\1 (cxccptional casc marking) vcrb which assilns Case to the

e.bcdd~d subjcct, \4r. Thin. in scntenrc (cl}.

~-

~- - ,,~-

~-i _ t< -;:-~ F :~~~~

-. -- - ,-'

~--

,-

-62-

1

1

T •

Cbapter 5

EXPERIMENTAL STUDY

This chapter reports on an experimental study of L2 acquisition

of English reflexives by Japanese ~peakers and investigatcs the

ideas developed in the previous chapters; namely, lhe t.heory of UG,

the Subset Principle and transfer wi th respect ta the GCP and the

PAP.

5.1. Rationale behind the Study and Two Hypatheses

The maIn concern of the study is ta investigale how learners

set the val ue of a parameter where the values of the LI (.Japanese)

and the L2 (English) differ. The two parameters ta be investigated

are the Governing Category Parame ter (GCP) for which the LI is

marked and the L2 i s unmarked, and lhe Proper Antecedent Parameter

(PAP) for which the LI is unmarked and lhe L2 is marked. If the

Subset Principle opcratcs. thcre should be no misinterpretation of

English refl cxi ves, as Engl ish has the most restricli ve sctli ng for

the Gep. 1 have hypothesi zed that the Subset Princi pl e wi 11 not

1

1

1

operate and that there may be p.rron; hecause of transfer of the LI

parameter value.

1 have also hypothesized that transfer from the LI may include

transfer of preferences. That is. even where the language allows

several possibilities regarding the antecedent of reflexives. it

is possible that native speakers have cl preference for one possible

antecedent over another. 1 f \t:arners transfer their preference in

Ll to L2. their L2 mistakcs \\hich are not overtly explainable by

the LI or the L2 may in faet be traceable to LI pref{~rences.

Even if \f'arnprs make transfer errors. there may be changes

over time. during the subjects' e'\posure to English. leading ta

acquisition of the L2 values. 10 ('nsure that the experiment \\ould

be sensi li ve to such progrcss. the subjects were selected from

different grade levels.

5.2. Method

5.2.1. Subjects

There were 6 groups of subjects involved in the experiment: 4

experimental groups (native speakers of Japanese learning English as

a second 1 anguage), a Japanese control group (nati ve speakers of

Japanese), and an English control group (nati ve speakers of

English ).

- 6 4 --

r

1

1

T

(ll~~pprtmt)ntal_ (~rgups

The t2\pt~rimt~ntcll groups consish'd of four levels of ~üulit:nts:

grade 10 (}1 subjt'cts, dgf~ F5·1b), grddp Il (H subjects. a~w lh­

l'n, grade 12 (13 ~ubjpcts, alV~ 17~13)' and grade 11 (20 subjerls.

age 18 ~19), Subjects in grade 10-12 \\l'r(' in the first tn lh(~ lhird

year of s~nior high sehonl and thost' in grade H \\ere first-year

students at college in ,Japan. The gradp 10,12 ~ubjects attcndcd d

privatc b ~t'ar sf~condary sehoul (gradt~ 7- 12) located in lbaraki.

,Japan, \\hcn~ mns\ of them beld startt:d It',lrning tnglish al the age of

12. when tllt'Y {'nt('n-~d lhb school. lht:y n~ceivcd rnglish tessons

ilpproximalch f> limc~ a \\Pek; cach class \\as fi minutf's lung. and al

least Ont' of lhe 6 rlasses \\as taught by a nat i Vp speaker of

English. At the lime of tcsting. Ulf~Y \\cre In the mosl advanced

level out of fuur in \;cll'h grade.

1 n t h(~ lexlbooks l!l<;se subjcct s lI~ed, English rcfl(:xi vcs were

introduced ln lhe heginning of t1H: second year. lIowever. no

e:<plicit instruction \\as given \\ith regard to the anlecedent of

reflexives 1.

Grade 13 (20 subjects. age 18~ 19) subjects were first- year

students at a college whirh is localed in Yokohama, Japan. They were

graduates from various secondary schools.

It was planncd to have subjects ~hosc backgrounds werc identical

as far as possible. Subjects were asked ta fill out a questionaire

before tht~ f~:\pf'rimfml. They \M~n~ to idf'ntify: the (lgf~ at \\hich they

started ~nglish. the amount of expo~ure to F.ngli~h. any experience

- 6 5-

1

(

abroad, and kno\\ledge of other languages besides l:nglish. Wlll:n the

data \\ere gathered. subjccts \\ho had had f>arly e'\posure to English

were eliminatecJ; thus, most subjects held started lt~arning English at

secondary ~('hool (age 12), ~hile sorne students started within a year

of entering secondary school (age Il). ~ost subjects reported that

they spend sorne tirne working on English through homework assignrnents

outside the classroorn. Those 'v\ho had lived outside of Japan were

excluded. Regarding kno'v\ ledge of othcr foreign 1 anguages. Grade 13

subjects kne\\ either French or German besides English, as a foreign

language course ottwr than English \\a~ n~quired at the ('ollege. No

one indicated that her knowledge of another language \\as superior to

that of f.nglish. There were a ft~'" subjects in Grade 10 ... 12 \\ho

indicated that they knew sorne other language. but again the level

~as far belo'v\ than that of Engl ish.

Before the actual test on reflexives, a Syntax Test was

administered, which was designed to f'\arninc whether or not subjects

had actually mastered the structures and the vocabulary that were

going to be tested. If they had not yet acquired the structures.

i t was considered irrelevant to test the binding of reflexi ves on

them. (This reasoning, based on Otsu (1981) and others, will be

explained below.) The 65 subjects ail passed the Syntax Test and

they were considered to be eligible for the UG Test.

169 students participated the experiment; however, on the basis

of the criteria described above, SI subjects were rejected because

of lheir experience abroad, 15 because of their early exposure to

-66-

1

i

English, and 38 berause they failed th(: Synld\ rt~St. 2

Grade 10 subjects actually came from two CL1SSt;~; howcver. as

there was no statistically ~ignificant difft:r!:Il('t: in rt'sponscs

between thcse t wo classes. tht: i r resu 1 t s werc comb i ncd. 1

In sumo the subjects in the expcrimental groups hdd had al

least 3 years (i n the case of Grade 10 subject s) to b years (in the

case of Grade 13 subjects) of formdl English instruction in Japan

when they \\ere testcd. They i\ere consider<:d to have mastered aIl the

structUrf:S and vocabulary that \\ould clPPCdr in the acludl test.

(ii~p.911_ese ControJ_Grou~

A Japanese control group ronsisled of 22 Grade 12 subjects (age

17~18). \\ho altendcd the Sdme scrondary school as the experimental

groups. They responded lo .]dpélnese versions of the lest sentences.

( i ii) Engl i~~~_ntr~LJl~Q!l~

20 college students (age lï~19) ln Montreal. Canada, served as

English controls. They \\f:re nati ve sp~:akers of Engli~h. but had

learned French as a second language (the length of their exposure ta

French ranged from 5 to 14 years), lhey were lcsted on the same

experimental test as the experimental groups.

5.2.2. Materials

The test was composed of l\\o parts: a Syntax Test and a UG

- 6 7--

,

1

Test. The Syntax Test was to ensure that subjects had mastered the

structures and the vocabulary in the LfJ Test. The UG Test \\as the

test which actuall y exami ned the acqu i si tion of I:ngl i sh reflexi yeso

The contents of both tests are described below.

5.2.2.1. Syntax Test

The Syntax Test included four sub-tests (the whole test 18

given in Appendix A.)

( i) Anaphora Test A

There were four sets of sentences \\hich \\ere designed to

examine \\hether subjects could correctly differentiate pronominals

and reflexi ves wi th respect to thei r antecedents 4. The sentences ,

were made up of simple clauses; if the subjects could not interpret

the antecedent of a reflexi ve in a simple clause sentence, i t meant

that they had not acquired the propertics of reflexi ves (\\hich must

be bound in their governing category) and pronominals (which must be

free in their governing category). Examples are as follows:

-68-

1

T

( 33) Al ice and June took a ba t.h. .Jllne washed her~el f.

'il "ST ION Who \\ ashed \\ ho 1)

.\l'N ... · 1:. 1< \\dshp.d ----- --

(34) John went hunting with Bob. By accident, John killed him.

QW,TION Who killed who by dccident?

"N~WLII killed

Following each set of sentences. then~ was a question in Japanese,

requiring the subjects ta identify \\110 did something ta who. They

\Vere asked to ans\\er with namp.s. (Whp.n filling out the underlined

parts. it \\dS emphasiL(~d thal lhey Sh()llld not use pronominals (e.g.

kare 'he', kanojo 'she') or n;flp.\lve~ (zibun 'self), which would

make the ans\\ers arnbi guous.)

For each set of sentences, the first introduced two NPs of the

same gender, in arder to avoid tilt: si tllation \\here subjects choose

the antecedent on the basis of the gendcr only. Test. sentences \\cre

designed t.o be of about the same lenglh and as semantic311y neutral

as possible in order to minirnize thf' possibility that sorne pragmatic

factor would favor one of the NPs. The second sentence in each set

was count.erbalanced as ta the subject NP. In some cases, the

subject NP equaled the NP \\hich appeared first in the first sentence

(e.g. John in the second sentence in (34) matches the fi rst NP in

the first sentence). In other cases, the subject NP equaled the

second NP in the first sentence (c.g . .June in the s(~cond sentence

in (33) matches the second NP in the first sentence). In this way,

-- 69-

1

1

it was hoped any subtle eues. e.g. oid vs new information, \\hich

might affect the interpretation of the antecedent would be

eliminated.

(ii) Anaphora Test B

Even when the subject interpreted the antecedent in Anaphora

Test A. there was still a possibility that the subject failed ta

understand it correctly. fhat is, for the set (32), the subject

might have allowed both NPs, ~une and Mary, ta be the antecedent of

herself. but was forced ta choose onl v one of them as an answer. Ta

avoid this possibility, Anaphora Test B \\as conducted." It was

intended ta determine whethp,r subjects allo\\ed only one passibility

for the reflexive and one for the pronominal. The subjects were

asked ta choose which of t \\0 sentences fo Il ows an i ni tial sentence

more naturally. There were four sets of sentences. Examples are:

(33) Bill often fights wi th Ann.

i. Bill al ways hurts him.

ii. Bill al ways hurts her.

(34) May and Paul taok a bath.

i. Paul washed himself.

iL Paul wa~hed herself.

This time. twa genders were involved in each set. If the subject

--70--

1

T

chose (33i), it means that ~im refers ta Bill. suggesting that the

subject interprets him ta be a rcflexive as \\ell as a pronominal.

If the subject chose (34ii), it means that herself refers ta May,

suggesting that the subject interprets herself to be a pronominal

as weIl as a reflexi ve. Such responses would indicate that the

subject doe~ not have command of the relevant distinctions in

English.

(Hi) Structure Test

Subjects were asked ta translate seven sentences from English

ta Japanese (see II in Appendi x A). This test was designed ta

examine whether the subjerts \\cre capable of understanding aIl the

structures which would appear on the UG Test. There were two-clause

sentences (one cl ause tensed and the other i nf i ni t i val. or both

tensed). three-clause sentences ( t wo clauses tensed and one

infini tivaI. or aIl three tensed). and one-clause sentences wi th a

subject. a direct object and an indirect object. Translations were

scored on the basis of understanding of the structures and not of

understanding af the meanings of words.

(iv) Vocabulary Test

The words which were used in the whole test were suppased ta be

familiar ta the subjects. Nevertheless. a sheet of paper with a

vocabulary list was handed ta the subjects 1 ... 3 day( s) before the

test day, in arder ta make sure that the subjects understood the

-71-

t

1

1

meaning of a11 the words. Just before the test, proper names were

reviewed sC) that there would be no confusion about the gender of

proper names. The vocabulary test included 10 ~ords; subjects were

asked to gi ve t.he meani ng of each ward in Japanese. 1 f the subject

could pass the structure and vocabulary tests, he or she was

considered to know aIl the words. However, there was one student in

each group who could not give the meaning or gave the wrong meaning

for the verb. blame. Si nce these subjects answered perfectly in

other tests. their responses for the sentences wi th blame in the UG

Test were not counted.

5.2.2.2. UG Test

The UG Test was a multiple-choice grammaticali ty judgement test

designed ta exami ne the subjects' i nterpretation of the reflexi ve

with respect to the GCP and the PAP. It consisted of five sentence

types. Each type was represented by five sentences so that a total

of 25 sentences were included in the test. The subjtcts aIl

received the sentences in the same order. It was an unpaced task:

however, subjects were encouraged not to spend too much time on each

item.

Written and oral instructions were given. 6 Subjects were

instructed that the purpose of the study was to find out how the

subjects feel about certairl English sentences, and not to test their

knowledge. so that they should relax and answer every question in

-72--

1

1

J

order, gi ving their ini tial response ta each sentence. They \\ere

also tald not to go back and change their answers. They ~ere asked

to indicate who himself or herself referred to in each sentence by

circling one of a set of given choices. For example,

(35) John said that Bi 11 hi t himself.

a. John

b. Bill

c. either John or Bill

d. someone else

e. don't know

If the sentence was ambiguous, they ~ere to choose an either A or B

type of response like (c); if they did not find the proper

antecedent in the choices, they were ta circle someone else and to

write down who it referred to in the underlined position. When they

did not understand the sentence, they \\erf' to circle don't know. The

reason that the someone else choice was included was that a

corresponding Japanese reflexive, zibun, might be interpreted as

having the speaker as i ts antecedent. 1 t was considered that the

subjects might use this strategy in English. Moreover, when several

NPs are possible antencedents for zibun in Japanese, zibun becomes

ambiguous, which was why choice such as ei ther John or Bill among

the multiple choices were included.

Sentence types A and B were made up of fini te clauses; while

-73-

•

1

1

Types C and D had an infini Uval clause in the most embedded

posi tion. Types A and C \\cre biclausa 1. Types Band 0 ",ere three­

c1ausal. and Type 1: ",as uniclausal. Type E had a subject NP. an

indirect abject NP. and a reflexi ve in a direct object posi tian.

Examples for these five types arc given belo~:

( 36)

Type A: h,o-clause Sentence (fini te)

John said that Bi 11 hi t hi msel f.

[NPl [NP2 refl. J J

Type B: Three-dause Sentence (fini te)

Mary remembers that June said that Alice blamed herself.

[NPl lNP2 lNP3 refl. ]]]

Type C: T\-\o-clause Sentence (infinitf')

Jllne wants May ta undersland hersel f.

[NPI [NP2 refl. ]]

Type D: lhree-clause Sentence (infinile)

Tom says that Paul told Bob ta introduce himself.

[NPl [NP2 [NP3 refl. ]J]

Type E: One-clause sentence

Bob talked ta Paul about himself.

[NPl =subj. NP2=obj. refl. 1]

For Types At C, E, fi ve choices were gi ven: NP!. NP2. ei ther NP! or

NP2. §gme_qne_' els~. and ~9_n·LJ5!:!Q~. These multiple choices were

presented vertically after each sentence. For example.

-74-

1 ( 37) John said that Bi 11 hit hi œsel f. ----_._-,-

a. John

b. Bi 11

c. ei ther John or Bill

d. someone else

e. don't know

For Types B and D, nI ne choices were gi ven: ~_Pl. ~P2, NP~, et ther

NP! or ~P.2, either !P2 or_JiPJ, eith~.t_ \P3 or NPl, eijJler NPl or NP2

or NP3, someune cise and don't. kno\\. For cxample,

(38) Tom says that Paul told Bob to introduce himsel!.

a. Tom

b. Paul

c. Bob

d. ei ther Tom or Paul

e. ei ther Paul or Bob ~ ~ ,.1_ __ ~ "-- _ ~_ __

~[ - - -- -~ - ~

f. ei ther Bob or Tom ~T _ - __ ..:' ___ :.... __ -.:- __ ~ __

" -

g. either Tom or Paul or Bob

h. someone el se ------,

don't know 1.

Multiple choices were given in Japanese (names were written in

English) for the experimentdl groups in order to avoid any

misunderstanding in choosing the answer.

-75-

1

l

,

English controis élnd ,Japanese conlrais rf'sponded ta the Sdme

sentences in Engl ish tlnd in .Japancse respect i vel y; furthermore, for

the t\\O control groups, th(~ don't kno\\ choiCf~ \\as omitted. since

they were respon~ing to questions about their native language.

In the CG Test, the Vf~rbs in (39) appeared in the ernbedded

clause \\ith a reflexive. and the rest of verbs used were given in

(40). These verbs were used in simple pasl, or simple present.

except for a few places \\here USf~ of v.iIl \\as more plausible.

(39) hi t, understand, hate, introduC'e. blame

(40) say, kno\\. think. believe, ff'member. tell. ask. want

5.3. Results

Resul ts of the experirnental gro'ups are given first in section

5.3.1. The cornparison of these groups \\'ith h\o controis is shown in

section 5.3.2, AlI the individual responses are found 10 Appendix

c.

5.3.1. Resul ts of the experirnental groups

The number and the percenlage of aIl the responses obtained

from the experimental groups are gi ven in Table A. They are broken

down by sentence type and grade.

-76-

1 Tnble ,,: The nl!.J1l~~ of--'.:C'sJ~)!~'-'~'-~ _l,n,l~~~h_ ,~YJ.)(\ hv.g.rJdp

Grade 10 GrJdc Il (Jr;)(k 12 Gr.1dl' 13 Total -Tn-=T3T- -Cô-::tt-) -(n~rHT -(i1=-lO)- [n;-G~)

fType Al ~PI

.. 18 1::; 1:::; 55 07.131) 1

W2 C;2 50 (18 77 247 (76.951 ) Wll2 1=) 1 (, - II) ( 1=).921) 1

don' t know 0 0 0 0 0 (total) 64 ---6<) --- ---8(1' C)<} 321

[Type Cl Wl 21 :'7 3i 28 117 (36.451 ) W2 32 10 16 59 177 (55. lU)

~Pl12 (, ") 5 12 25 ( 7. 7C)1) H

Jon' l know 1 0 1 a ") ( 0.(21) .. (total T- 61 6f)

. . ------gq .- , c)c) 321

[Type Bl Wl "3 5 -l 1 13 ( 1.0:::;%) ... W2 11 11 1G 20 61 ( 1 C). 00% ) \P3 H ,12 (,~ 68 217 (6i.60%)

WI12 0 (, a 1 la ( 3.12%) ~P2/3 1 1 1 '5 Il ( 3. B%) ~Pl/3 1 0 1 0 ") ( 0.62%) ...

WII2I3 2 1 ") 0 5 ( 1. S(1) ... don' t know 2 0 a 0 ") ( 0.62%) ...

(tolal) M (1) Wl qC) 321

[Type D) Wl 4 -1 2 2 12 ( 3 741) :\P2 If) 30 32 v ... 0 107 03.33% ) ~P3 31=) 34 45 ~8 1- ") ,- (S3. 58%)

~P1/2 0 a 1 1 2 ( 0.62%) W2/3 3 1

,.. Il 22 ( 6.85%) 1

\PI/3 0 0 2 1 3 ( 0.93%) WI/2/3 2 0 0 0 2 ( 0.ü21)

don' t know 1 0 a 0 1 ( 0.311) (tolal) 64 69 8C) 99 321

[Type E] NPI 54 51 6<) 66 240 (73.851) ~P2 10 17 15 24 66 (20.311 )

WI/2 1 1 (, 10 18 ( 5.541) someone cise a 1 a a 1 ( 0.301)

( lolal) 65 70 90 ,\00 325

• ln lype A-O. thcrc were no rcsponscs for somcone cise. and in type E. ------- --thcre wcrc no responses for don' l know. •• One rcsponsc for someonc eise lndicillcd that il rcfcrred to a speaker. ".In lype A-O. therc-was--on~ subJect ln cilch grade whosc responscs for the sentence with vC'rb hl~m(' have hecn nmlttrrl

-11-

t

1

1

As for Types A --D. the most frequent response was the correct

one, Le. the 1 oca 1 antecedent, wh ich i s ~P2 in Types A and C, and

NP3 in Types Band D; ho~ever, lhere \\ere subje:::ts who chose the

wrong antecedent for the refl exi ve. i.e. the non-local antecedent or

the ambiguous antecedents (1 ~ill u~e ~mbigu~~ __ ~ntece~~t~ ta mean

aIl the responses \\hich allowed more than one antecedent. for

example, eit.her NPI or NP2, ~ithf'!" __ ~Pl or NP2 gr ~fl etc.). These

errors are evidence for the non-operation of the Subset Principle

which predicts only local antecedents for the reflexive.

Table B below shows the number and the percentage of local

responses and non local responses collapsing ail the ambiguous

antecedents wi th the non-local ones.

-78---

1 Table B: Local RffipJ~~_~~_ 'oQ..::..!9calJJ~~.!\--qes

Grade 10 Gradp Il Gradl' 12 Grade 13 Total -- -- ~- - - -~ ------ - ------

(n=13) (n"'U) (n-·18) (n<~O) (n-65)

[Type AJ

Local 52 sa 68 il 347 (76.95%)

Non-local 12 19 21 'J) 74 (B. 05%) ...... (total) 64 69 8Q 99 321

[Type Cl

Local 32 -tO .1.6 59 177 (55. 14%)

Non-local 30 :!) 42 40 142 (44.24%)

don' t know 1 0 1 0 2 ( 0.62%) -------~ --------

(lolal) 64 69 89 9e) 321

[Type B]

Local 44 42 63 68 217 (67.60%)

Non-local 20 Z7 :!h 31 104 (31. 78%)

don' t know ') 0 0 0 2 ( 0.62%) '"" (total) 64 69 89 99 321

[Type D]

Local 35 34 45 58 172 (53.58%)

Non-local 45 35 44 41 149 (46.11%)

don't know 1 0 0 0 1 ( 0.31%)

(total) 64 69 89 99 321

+Local rESponses are correct and non-local rES{XJnses are incorrect in

, aIl the cases.

-79-

1

"'f

Ta examine "hether or not there dre any differences in grades

and in types. the mean number of the correct responses in Types A ""0

for each group of subjects \\as calculated. The maximum possible

score is S for each type.

Tab} e C: Mean Number of the Correct Responses (by type and gracIe)

Typ~ Type ~ Type C Type 0

Grade 10 4.00 3.38 2.·16 2.69

Grade Il 3.57 3.00 2.86 2.43

Grade 12 3.78 3.50 2.55 ? -0 ..... ..,

Grade 13 3.85 3.40 2.95 2.90

mean 3.800 3.338 ,.., ""')3 _.1 .. 2.646

Analysis of variance shows that there is no significant grade effect

(F(l,3)=.17 pc.918), nor interact ion of Jrade by t~pe effect

(F(9.183)=.S5 p:'-.339). However, therc is significant effect of

typ~ (a multivariate test of significance shows F( 1.3)= 13.766

peOOO). Thus there are signi ficant di fferences in responses among

the four types. but nr)t among the four grades.

This result, i.e. that there 1S no tendency toward more

accurate Interpretation of reflexi ves wi th i ncreasi ng grade, is

somewhat surprising, as 1 had expected that the higher the grade

was, the better the subject's proficiency \\ould become. (This point

will be returned to below.)

-80-

1 As there is no difference heh\f'Πgradt;s, 1 dccided to collapse

the results of the four grades inta ont: t:\perimental group. When

the results are takt'n together. !':)ignif icant differences are found in

the follo\\ing pairs (A Tukey's HSD tpst was performed: '\I!>l!14. 183)=

.472 p(.05): Types A & B UH,>.) ..1(2), A & C UlIsu.:l.077), A & D

(;"5u-=1.154). B & C U"su- .615), and U & D U llw ::c.6<J2). Therefore.

only the difference between lypes C & D is not significant ('"SU=

.077) .

The subjects die! best ln T\p<, A sentences (mean 3.800),

followed by Type R (mean 1.333). Tllf' SUhjf:Cts performed the \\orst

in Types C and D spnt('nCf~s \\hf:re no difference is found. Regarding

only the t\\O clause structures. tht~ 1.2 It":drncrs I\ere much more

accurate in finite clause sentences (l:pl~ A) than in infinite clause

sentences (Type C). They i nterpl~d('d more non-local ~Ps as the

antecedent of the reflexive ln hpe C than in Type A. which

replicates Finer and Broselow's finding (1986). This 1S an

interesting rt:sult and 1 will return to Il belm\'.

When the sentences were made up of t hree clauses as in Type B

and Type D. the subjects tended ta make more non- local choices.

They were less accurate in Type B than in Type A. al though both

structures had only fini te clauses. This suggests that the complex

structure of Type B had an eHect on subjects' identification of the

correct antecedent. somehow making the reflexive more ambiguous.

However, an interesting compari~on here is that between Type C

and Type D. Type C is two-clau~e structure white Type D is three-

-81-

clause structure, both of thern including infinitivals. ln contrast

to Types A and B. the cOllplexÏty of the structure does not seem ta

have an effect; i.e. there is no significant difference between

these structures. Moreover, the subjects chose local antecedents

more on Type B, with a three-clause tensed structure, than on Type

C, with a two-clause infinilival structure, su&gesting that the

subjects were affected by the infini ti val more than by the levels of

embedding. The question anses as to what the non-significant

difference betwCf.l types C and [) really means.

Th( results from Type [: ~hO\\s that the subjects chose subject

antecedents (71.35% of the time) more often than abject antecedents

(20.31%). Very few subjects chose the response of ambiguous (5.54%)

and only one subject bound the refl exi ve to the speaker. Thus we

can conclude that these subjects had a preference for subject

antecedents avcr abject antecedents and did not often see the

reflexive as ambiguous. contrary ta one of the predictions of the

PAP theory.

5.3.2. Experimental Group vs English ContraIs and Japanese Contrais

The number and the percent age of the responses for the whole

test from the three different groups are gi ven in Table D. (The

chaices of dan't know and someone else have been removed.)

-82-

Table D: Overall Responses of Each Group

Control (Engl ish) L2 leamers Control (Japanese)

[Type A] NPI 1 55 07.13%) 69 (62.73%) NP2 <J) 247 (76.93%) 'E (:J>. 36%)

NPI/2 0 19 ( 5.92%) 11 ( 9.IOX) 100 321 109

[Type C] NP1 2 117 (36.45%) 78 (70.91%) NP2 98 177 (55. 141) 21 (19.09X)

NPI/2 0 :5 ( 7.79%) 11 OO.OOX) 100 319 110

[Type B] NP1 1 13 ( 4.{f)1) 19 07. :..'7X) NP2 0 61 (19. (01) 58 (52.73%) NP3 98 217 (67.601) 10 ( 9.09X)

NP1/2 0 10 ( 3.12%) 5 ( 4.55%)

1 NP2I3 1 11 ( 3.43%) 13 (11.82%) NP1/3 0 2 ( 0.62%) o ( 0%)

NP1/213 0 5 ( 1. 56%) 5 ( 4.55%) 100 :319 110

[Type D] NP1 1 12 ( 3.74%) 14 (12.73%) NP2 1 107 (33.33%) Eh (60.00%) NP3 98 172 (53.58%) 12 (10.91%)

NP1/2 0 2 ( 0.62%) 8 ( 7. 'lJ%) NP2I3 0 22 ( 6.85%) 5 ( 4.55%) NP1/3 0 3 ( 0.93%) 2 ( 1.82%)

NPl/213 0 2 ( 0.62%) 3 ( 2.731) 100 320 110

[Type E] NPI 67 240 (73.85%) 105 (95.45%) NP2 21 Eh (20.311) 2 ( 1.82%)

NPl/2 12 18 ( 5.541) 3 ( 2.731) 100 323 110

" +Percentages have been rounded off to the second decimal place.

-83-

1

1

1 will consider these resul ts in t'ha parts: Types A'D with

respect ta the Gep. and Type E 'h i th respect ta the PAP.

5.3.2.1. Type A-Type D

English contraIs overwheImingIy chose the local antecedents

(98~99%), consistent with the theory of the GCP.

Japanese controls showed a defini te preference for the non­

local antecedent over the local one. In Types A and C "here there

were two possible antecedents (either a local or il non-local NP).

there were more subjects who chose the non-local antecedent (62.73%

in Type A and 70.91% in Type C) than those who chose the local

antecedent (26.36% in Type A and 19.09% in Type C). In both types,

about 10% of the responses indicated more than one possible

antecedent. As Japanese is the most inclusi ve language wi th respect

ta the GCP. any NP can be the antecedent for the rcflexive in these

sentences. l t follows, then. that \\c could expect many subjects ta

notice this ambigui ty. However. there were not many responses 10

each type which indicated that more than one antecedent is possible.

It may be that native speakers (and Iearners) simply notice one

interpreta tian even though others are available. 1 fit is the case

that native speakers of Japanese do not notice ambiguity where there

actually is ambiguity, we might expect the local antecedent and the

non-local antecedent ta be chosen al an equally frequent rate.

However. the non-local antecedent was chosen much more freQuently

-84-

1

J

than the local antccedŒt, suggesting that there \\as d preference

for the non-local antecedenl over the local antecedent among native

speakers even though the grammar ail m'vs ho possi bl e antecedents.

When there were three possible antecedents (Types Band D),

the middlc- NP was chosen most frequentl y (52.73% in Type Band

60.00% in Type D), follo\\cd by the main clause NP 07.27% in Type B

and 12.73% in Type DL The local NPs were chosen least freQuently

(9.09% in Type Band 10.91% in Type DL In bath cases, there were

sorne subjects \\ho found dmbiguily in interpreting the antecedent;

20.91% in Type B, and 16.36% in Type D. Among these subjects, 4,55%

for Type Band 2.73% for Type D responded with eilher NPI or NP2 or

NP3. The rernainder indicated that there were two possible

antecedents.

In the Japanese control group, there IS no significant

difference in responses between Types A and C (1 2 ::2.41 p >.30) nor

between Types Band D ([2-0.075 P ).99), Thus the Question again

arises as ta \\Ihy the L2 learners behaved differently \\Ii th respect to

Type A and Type C.

The experimental group's responses are distinct from those of

both the English controls and the Japanese contraIs. That is. these

L2 learners did not arrive at the correct setting of the GCP

parameter; moreover. their response pattern did not match that of

the Japanese contraIs. The resul ts are problematic for the Subset

Principle hypothesis. in that the learners did not start from the

most unmarked. smallest grammar, bu t set the value wider than i t

-85-

t

1

should be. allowing non-local antecedents in a number of cases.

Although the experimental group and the Japanese controls gi ve

different response patterns. 1 will argue (in section 5.4.) that the

learners still have the widest value of the Gep.

5.3.2.2. Type E

The Type E structure was included in arder ta examine the PAP.

This type of sentence had a subject NP, an indirect abject NP and a

direct object NP containing a reflexive.

(41) John showeù Bill fi picture of himseif.

In English, there are two possible antececlents in such

sentences since the grammar has the superset value of the PAP; on

the other hand, there is only one possible antecedent in Japanese as

it has the subset value of the parame ter. As Table D shows, the

Japanese control group chose the subject antecedent almost

exclushely (95.45X), consistent wi th what the theory predicts.

Where English controis are concerned. i t seems again that there is a

preference for the subject antecedent among native speakers of

English, even though the theory predicts that ei ther a subject or an

object NP may be the antecedent. Native speakers of English bound

the reflexive ta the subject NP in 67% of the sentences, ta the

abject NP in 21%, and to bath NPs in 12%. reflecting the ambiguity.

-86-

1 This result matches that of Thomas (1933).

The experimental group, on the \\hoIe, behaved more like native

speakers of English than like native speakers of Japanese. In fact,

there is no significant difference between the English controls and

the L2 learners (1 2 =11.096 P >.05); on the other hand, thcre is ~

significant difference between the Japanese controls and the L2

learners (1 2 =23.950 p <'00l). This result suggests that the L2

learners have already set the value of the PAP ta the superset of

English, al though many preferred the subject over the object

antecedent as nati ve speakers did.

5.3.3. Preference vs Pragmatic factors

As anticipated, the results show that native speakers have a

preference for one element over another when there are several

possibili ties which are allowed by the grammar. That is, English

contraIs showed a preference for subject antecedents over object

antecedents; Japanese contraIs revealed preference for non-local

antecedents over local antecedents, but this non-local antecedent

was not the one farthest from the reflexi ve as is shawn in their

judgements for Types Band D. In section 4.3.. 1 suggested that

this kind of preference might be due ta pragmatic factors,

especially ln Japanese. If such a preference is determined

pragmatically. we may find evidence that Japanese native speakers

bind the reflexive to the antecedent quite differently depending on

-81--

1: the verb in each sen tence. Berc would like to examine the

responses of Japanesc controls brokcn down by lype and verb. Five

verbs ( hi t. unders land. i nt roduce. blamc) are used ln the

clause conlaining a reflexi ve in Type A through Type D. The

experimental group's responses are also shown graphically in Figure

A.

~ --~--~1~:{~;~~~~~-----~ _ n _-~'_'---- _

__ ~~ __ r.---==-_'=;;:'~ _-

__ ~;~+~7~:i_1i~

-,

-.. -

, ,

-" -,..- \~

-88-

(

f

Figure A: Non-local Respot1SES in Tw A and Type C

(by Japanese controls and L2 leamers re=;pecti vely)

100

<.Xl

80

70

60

50

40

30

20

10

..... , ... , .. -,.. , ......... ..- , ---_._ .. .....;. '~....:. _.. ' .. • • • • .

• • • • • ',TYPE C (LI)

TYPE A (LI)

A (1.2)

O~--~------~--------------~------~---hi t understand hate introduce blélJœ

Figure B: Non-local Responses in Type B and Type D

(by Japanese controis and L2 leamers respecli vely)

100

ex>

80

70

60

50

40

30

20

10

.............•....•

.......... .. -

----_._._._ .•. _._._._._._-_._ ... -.. .........

.. ' .. '

o (La)

TYPE B (LZ)

O~--~----~------~------~------~----hi t understand ha te introduce blame

-8 g--

1

,

..\mong Japclnese conlrllis. \\/H'n t ht' t~mbf~ddf~d V<'rb was hi t, <)4.45%

(TYPt~ A) dnd ~(J.1b% (rVpt' C) uf l/w sllb.i(~cts chose non·· local

anlecedf'nts, whik in scnb~r1('('s wJlIl litt' v('rb hlame', only SO.OO%

(Typ<" A) dnd S9.0C)% (TyP(~ C) of the subjccts (hosc non-local

anlccedents. T11(~ non local prdt'rt'IH't' pattern is identical in Types

A and C, e\('cpt for the \t~rb introdIHt'. HI'sponse péllU~rs in Types B

and 0 are sho\\n in rigun~ B. Ihe' patterns in Type B and Type D do

not quiU' match those in lvpe A dlHI Type C respectively, although it

might be s.lid filai local judgements inCrf'élSC wilh t.I\f~ verb blame ln

Types A t hrough D.

dlso suggestcd that wh(~n there IS il strong prefcnmce for the

non -1 oca 1 dn t cceden tin lhe na li ve 1 anguage, lhere mi gh t be more

mis takcs in 1.2 judgcment; i.(~. more non -1 oca 1 antecedent

interpretalions. However, it is nol at ail clcar that lhis is the

case. The verb hLt in Type A gets the most non-local judgements

from Japanese controls but mostly local judgemenls in the L2. The

verb intr9~!lce in Type C gels the mosl local judgements in the L2,

in contrasl ta the LI. Similarly, understand in Type B gets the

most local judgements in the L2 but it gets relatively more

non -1 oca 1 judgemen ts in the LI.

1 n sum, although the non -local judgements of L2 learners

clearly differ from sentence ta sentence, this result cannat be

regarded as resul ting from a di rect transfer of their L1

preferences, failing ta support my hypothesis of a direct influence

of preferences in the LIon the L2.

-90-

a

1 5.4. Discussion

As the above resul ts show, we have obtained evidence that the

Subset Principle does not operate in L2 acquisition. Our L2

learners fail to set the value of the Gep correct Iy. They set the

value wider t.han it should be, allo\\<ing non··local antecedents for

the reflexive even in tensed clauses. This resul t is consbtent

with that of Thomas (1988). Reflexives in infinitival clauses (Type

C) recei ved more non-local responses. This resul t is consistent

with that reported by Finer and Broselow (19Rh).

Finer and Broselow suggesl that learners set the GCP ta an

intermedi ate val ue, dis ti nct from ei ther their LI or L2. As Finer

and Broselow's subjects correctly judged Type A sentences (91.7%) to

have local antecedents but were much less accurate on Type C

sentences (58.3%), their explanation holds for their subjects. A

study by Finer and Broselow (1989) replicated this result with many

more subjects. However. my subjects made a considerably larger

number of mistakes in Type A sentences (2,.05% of the cases, 17.13%

being non-local, and 5.92% being ei ther non-local or local). This

result is inconsistent with the value they assume since no non-local

responses are predicted wi th tensed clauses. In order to account

for the non-local responses of rny subjects. i t is necessary to

assume that they have in fact adopted the widest value of the Gep,

Le. the value required by their L 1. This accounts for the

non-local responses in ail four sentence types (Types A~D).

-91-

1 When two- and lhrec clallst~ t('nst'd st'nlcnccs an~ cornpared. the

subjects made il ~n'éùer numb(~r of \'rrors in the more complex

sentenn'!-'; I.t~. they \\t're less dCl'Ur,lk ln ,ldopt inp; lhe 10cal-

antecedenl ln three-clause struclure. where thn'f~ possible

an teced('nls an~ i nvo 1 vcd. nWSt~ tTrOrs arp f'xpl a i ned if we assume

that the subjects lrJnsfern~d lhcir 1.1 p,H"amder sctting. If the

subjects 'o,\t.:re picking an interm~~dic1t(' value of lhe parameler. as

Finer alld Braselow suggcst.. then tht'y ::,I!ould not make errors like

choasing non local antcccdtmts or dmbiguous n~sponscs in the lensed

clauses. In lypf' B (and lypc D a~ \\1'11), only onc out of the seven

choices \\as grrlmmatical in r.nglish; rwvcrthelcss. ail responses are

possible according to the LI sctting of th(~ paramct{~r. lhe fact that -

NPl (the farlhest NP from lhe n~flexivc) \\élS choscn less frequently

than NP2 in Types B imd D by .Japallf~St~ cont.rols may be due lo their

pref erenccs among the seven poss i b il i tics. The L2 karners also

chose NPI murh less frequently (:1.05% in Type B. 3.74% in Type D),

suggesting sorne influence of lheir Il preferences here.

However. what rcrnains a rnystery if t.hey hav(~ in fact retained

the widest setting is that the learners made significantly more

errors in Type C sen tences than in Type A; i.e. the [ ± tensed] cl ause

distinction observed by Finer and Broselow has real effects. at

least in b~o-clause sentences. This difference cannat be attributed

ta the learners' native language. The sentence (42) is a Type C

sentence in Japanese.

-92-

1

(42) John-ga Bill-ni zibun-o shokaisuru yoni tanon-da.

John-nom Bill-dat sel f -acc introduce comp ask-past

'John asked Bill ta introduce SELF:

Although Bi 11 is case marked as an \\ndirect object by ni. Bi 11 cao

be the antecedent of zibun. This fact was explained by assuming the

following structure for this sentence (section 3.4.2. (18d'».

(43)

IP

NP~I' ~ ~

John ,-ga VP l ~

NP CP V -da' p~\St' ~ /

BilL-ni C' tanon-

IP./"\..C

NP ,/'\I' . yom

~J VP'I NP~V ~

zi bun 1/,-0 shokai suru

Since Bill contraIs the embedded subject PRO, it can be the

antecedent of zibun. Although 1 leave the issue of whether or not

Japanese has infinitivals open, this structure is similar ta

English. (44) shows the structure for '.John asked Bill ta intraduce

-93-

1

1

himself'.

(44)

IP

NP~I' 6 ~ John 1 VP

'past' v~ CP c6 /

ask Bill, C ~

C IP

NP~I' ~ f'....... PRO J 1 VP

ta V~NP ~

introduce himself J

ln bath (43) and (44), lherefore, Bill IS coindexed with the

reflexi ve in the embedded abject posi tian via PRO. The antecedent

of the reflexive is clearly decided structurally.

There is thus no evidence from Japanese or English that tells

the L2 learners that an English sentence with an infinitive clause

like (44) should be treated di fferenUy than a sentence like (43).

Fi ner and Brase 1 ow (1986) propose as a second poss i b i li ty tha t

learners interpret the infini ti val sentences by concentrating on

surface ward arder, sa that Bill IS cansidered an object and lS

avoided as the antecedent because of the PAP. If lhis is correct.

-94-

1

1

1

my L2 learners should have behavf:d more like Japane~e controls in

Type E sentences. That is, the PAP states that n;flexives in Type E

sentences can only be cart-~fcrent. \\lth !:.ubjects ir. .Jdpanese: if these

learners maintain the Japanese value of the PAP in English, they

should avoid abjects in English Typf: f', senlences as the i1ntecedent

of the reflexive. But in fact. they appear to have chosen the wider

setting of the PAP since they bound rcflcxives to objects in more

than 20% of the cases; ln contrast. .Japanese controls bound

reflexives almost entirely lo subjcrls (ovcr 9S% of lhe cases).

Thus, the LJ learners do not secm ta have dcalt 'hilh the infinitival

sentences on the basis of surface lpvel fcatures alone, contrary ta

Fi ner and Brosc 1 0 w 's proposa 1.

General!y speaking, thcrc were more subjects \\ho chose correct

antecedents than wrong antecedenls. 1 wou!d like ta emphasize this

point and argue that sorne subjects have set the correct value of the

parameter for English. For example, t.here "ere 10 subjects who

responded 100% correct.ly (subjects #5, 6, 7. 16. 22. JS. 40, 49, 50,

65; see Appendix C)7. These subjects show that resetting of the

parameter in the L2 lS possible and argue against the non-operation

of UG hypothesis as proposed by Shachter (1988 a,b) and Bley-Vroman

(in press). That is, these learners \\ere able to reset the value of

the GCP, from the widest val ue to the smallest val ue, even though

the LI grammar and the L2 input did not give them clear evidence for

this. There were also 6 subjects who responded almost perfectly but

made one error (subjects #8, 30. 32. 44. 45, 53). These subjects

-95--

may have bt'cn III i sI pd by lhf'i r LI in S(JIllC cases al though they \\cre in

the prorcss of arriving al tht: f'Orrt'l! L2 setting. Othcr learners.

however. ha\'t' cln i nappropricl t (' sd t i ng. Viht~n the st.ruc ture of lhe

L2 sentences included three emht~d(h~d clauses or an infinitival

clause, they lended ta have difficully in finding the correct

antecedenl. At Ihis moment. i t is not al all clear why Type C

scntences mi sied the subjccts more oft~n than Type A sentences.

assuming that hner and Broselow's n.planation is inappropriate for

thc!:>c slIbjt:ds. lhe n~sult also shO\\s lhat lhree-clau~;e tensed

sentences w<:re more difficul t than t \\'0 clause tensed sentences.

When resul ts for Type C élnd rYPt~ [) St~nlf~nces are comparcd. the

subjects did worse on Type C thdn on [ype D. suggesting lhat lhe

infinitival \\as the crucial factor kading ta difficulty in finding

the antecedent. 1I

A final question still remall1S. namely the lack of improvement

over the different grade levels !hat. tested. The subjects

reported in this thcsis are considcrcd ta bc low level English

learners, as they have rcccived English instruction only for three

to six years in a furmal classroom si tuation in Japan. and most of

them failed ta set the correct value of lhe GCP. Assuming that

Finer and Broselow's subjccts were more advanced (in that they were

exposed to English in the United Slates). it may be argued that

learners move from the widest value ta the narrower values as they

become more proficient in English. Thus 1 would like tu propose the

following stages for acquisi tion of English reflexi ves by L2

--96-

1

1

1

learners "hen their LI value is the "idest and the L2 value Is the

smallest:

I. Low-Ievel learners: They may fail to set the correct smallest

value because of transfer of the LI value. They set the widest

value; consequently, they make incorrect non-local antecedent

choices.

II. Intermediate learners: They do not trcat t.he L2 like their LI

nor give i t the correct L2 value. but choose an intermediate

value. They make a distinction between tensed and non-tensed

clauses, so tltat in infinitival clauses, they tend ta choose

non-local antecedents, but in tensed clauses, they choose local

antecedents correctly.

II 1. Advanced learners: They set the correct val ue for English.

They correctly bind the reflexiv~ to local antecedents whether

a clause is tensed or infinitival.

Obviously, this proposed sequence must be subjected to further

empirical investigation.

-97--

1

l ok

1

Footnotcs tü Chapter ~:

Tex t b ù 0 k sus c d (f u r g rad c s 7 ~ l) ) W l' r e . N a k êl m Il r a c t a 1. (' d s . 1 980 .

Thc Ncw_~r:ow_'!_J_~gl!..::>Jl_~~'.!!..,?~. 13 Tokyo: Sanscldo. Reflexive pronouns

appeared only in a f<":w pldres with êl simple dausc: e.g , "lt's Humpty

Dumpty," ~did AliCf' ta h('rsl'If. (S~'rJc~ 2.:S)', 'You ran starl talking about

yoursclf tu jour frlcnd. (SCrlf'''> 2.ïO)' (jrdoc 10-12 sludents usrd: Shimizu

ct al. cds. ICJ82 Th.: J!1J:~!".n.'!J.I.2!1~L .F_n_~I..!.~~._·_~~!~in_g Tokyo, J(altakusya.

and Saïto et al cds lC)gS B1Rh!Oi)rL.T~ J: . .n81ish~. J JJ Tokyo, Sanseido.

Non~~ of the t('\tbook~ IIdd ,'\plicii in"otru( lions with rCj1,drd to syntactic

propcrtlc~ of EngllSh rdlç;xlvc. dnd non\' of the tearhers al the schoül

indicated that thi'Y taught s\lch proprrties

2. A table bc10w :"1l0WS the classification nf the suhjr,cts invo\ved in the

expcrimcnt: # of the !'-.ubJC'rts in cach da~~. who held becn abroad, who

succccded the Syntax Test (hut had cdrly English exposure), who failed the

Syntax Tes t, respec t i ve 1 y.

never abroad

ta ta 1 abroad succeed(earlyexp) fa i 1

Grade 10 (1) 35 18 c:; ( 2) 10

( 2) 29 12 3 (2 ) 7

Grade Il .1} Il }.1 ( 3 ) 13

Grade 12 38 8 18 ( 4 ) 8

Grade 13 26 2 20 ( 4 ) 0

( total) 169 51 65 (15) 38

3. 1 2 tests wcre pcrfarmcd on each item. There wcrc no significant

diffcrcnces bctwcen the rcsponses of two classes at p = .05 levcl.

4. Originally, there were six test sentences. However. two sentences

turned out ra bc problematic and thcy have becn climinatcd. They wcre:

-98-

1

1 il

a. Bill fights wlth Paul Paul hurt!'> hlmself.

b. Bill teaches Paul how t.o play tennis Bill is proud of him.

For (a), several subjcrts illdlcatcd that 'Bill hurt Paul'. Although the

sen tencc was des 1 gned to re mo vc ~omc prag mat 1 c f ac tors, sorne subjcc ts scc rn cd

ta be confused as ta why Paul could hurt hirnsclf, sincc Paul was fighting

with Bill. Bill must have becn involvcd at lcast indircctly \tany subjccts

also made rnistakes on (b) Although it is not clear why thcy dld so, the

sentence was rernovcd

5. 1 am gratcful to Prof. Yukio Otsu who suggestcd to mc such a possibility

and thc type of tests to examine' iL

6. The data from the three cXPcrlmcntal groups (grade 10-12) and the

Japanese control group were gathcred at \1eikei High School. Ibaraki. Japan,

in July 1988, whcre the cxpcrirncnts wcrc conduded by the duthor. The data

from the one experlmental group (grade 13) werc gathercd at College of

Foreign StudlCS, Yokohama, Japan, ln July 1988, while thosc from the English

control group at LaSalle College, ~ontrea1. Canada, ln October 1988; both of

the cxperimcnts wcre administcred by the tcachers of the schoo1.

Î. Rccently, several people have ralscd questions about the binding of

reflcxives and hav~ made ncw proposaIs (c.g. Pica, 1987; Huang and Tang,

1988. etc.). Pica (1987) argues that therc arc two klnds of rcflexives: one

is XO. likc zibun. Vihich can be long-distance bound because it moves

eventua II y to 1 NFL through hcad ta he ad rnovement. the other i s X - -. 1 i ke

himsclf or karezisin 'himsclf' and zibunzlsin 'self-self' in Japanese, which

must be locally bound.

If Pica 1S correct and my subjects treated himself or herself as

Japanese karezisin and not zibun. this could explain correct choices of

local anteccdcnts. In facto subjccts were asked ta givc translations for

two test sentences at the end of the experiment. The translations for

Engllsh rcflexives were varlOus. for example, ~ibun, zibunzisin. karezisin,

proper names etc. Moreover, i t is not neccssari Iy the case that those

-99-

1 subjccts who put ~arczisin 0'- zibunzisin respondcd rorrcctly. Translations

of the subjects who perfor-mcd pcrfectly in English includcd zibun. Thus,

althaugh thesc differenres in the bchavior of Japanese reflexives rnight have

becn a confounding factor, they do nat appcar to explain cither the

succcssf ul or the unsuccessful 1 carncrs. nor the fini te Inoo-f i ni te

distinction.

8. There is sorne evidence which shows that the subjects had difficul ty in

infinitivals: out of 38 subjects who failed in the Syntax Test, 6 subjects

cauld not understand the structure of the sentence wi th the verb want, Il

subjccts wi th the vcrb ask, and 2 sub.iects w i th both verbs. Thus. 19 out of

38 subjccts failcd the Syntax Test because they made errors with these two

scn tences.

-100-

1

{

Chapter 6

CONCLUSION

A number of issues have been raised in this thesis with regard

to the operation of parameters of UG in L2 acquisition.

The experi mental study reported on here suggests that L2

learners transferred their LI parame ter setting, and consequently

made errors in the choice of antecedents for reflexives as predicted

by the transfer hypothesis. Thus we can conclude that the Subset

Principle did not operate properly in L2 acquisition and that

Hypothes i s 1 (secti on 4.4.) mus t be abandoned. Errors made by my

subjects varied from sentence t)'pe to sentence type; errors occured

most frequently in sentences with infinitival clauses in two- and

three-clause structures. followed by three-clause tensed sentences.

As the subjects chose a relatively high numher of non-local

antecedents in tensed clause sentences, Hypothesi s III. which states

that L2 learners choose an intermediate value. must also be

rejected. AIl the errors made by my subjects are explained if we

assume that they transferred their LI value for the Gep. Thus. the

resul ts suggest that Hypothesis II is correct. 1 t should be

-101-

1

1

,

emphasized that my results argue against Hypothesis V, the anti -UG

hypothesis. None of the subjects' responses was incompati bl e wi th a

grammar of an existing language.

Regarding Hypothesis 1 V, which predicts that learners transfer

preferences from the LI. the resul ts obtained here do not strongly

support this hypothesis. Al though nati ve speakers of Japanese have

a preference for non-local antecedents over local antecedents, and

this preference varies depending on the verb in the clause

containing the reflexive. the L2 learners do not seem to have been

affected directly by their L1 preference. If, for example. these

subjects had dealt with English sentences by translating them into

Japanese in each case, their responses might have shawn the same

pattern in choosing non-local antecedents as the one observed in the

Japanese control group's responses. Since this was not the case, it

is plausible ta conclude that the learners treated English

differently from Japanese. Thus, although 1 hypothesized that LI

transfer \\ould include tr~ansfer of LI preferences, the results do

not strongly support this view. Howevcr, the fact that the subjects

chose only a few farthest non-local antecedents in three-clause

sentences (Types Band D) may be regarded as a reflection of their

LI preference since there were not many farthest non-local

antecedents responses by Japanese controis.

When the LI yields the widest value of a parameter of UG and

the L2 the smallest, as is the case with the Gep, resetting appears

to be difficult and it takes a relatively long period: many Grade 13

-} 02--

1

1

subjects had not reset the parame ter. Nevertheless, resetting of

the parameter is possible. as 10 out of 65 subjects in this study

correctly bound the reflexi ve ta local antecedents in aIl the test

sentences. As far as 1 have been able ta establish. the behavior of

reflexives, Le. that English reflexives must be bound in their

governing category (the smallest domain), is not tausht in class.

Thus 1 would like to suggest that resetting of the parame ter is

possible despi te the absence of evi dence, suggesting the operation

of UG.

Wi th respect to the PAP, the native speakers of English are

found ta have a preference for subject antecedents over abject

antecedents. The resetting of this parame ter seems ta have been

easier for the learners, prompted, probably, by posi ti ve evidence.

since they were going from the smallest to the widest setting.

Their judgements were very similar to English contraIs.

Al though learners do not attain the appropriate L2 value of

parameters immediately, suggesting the non-operation of the Subset

Principle and the influence of the LI value, pararneter resetting

appears to be possible, at least for sorne learners.

-103--

1

..

APPENDIX A: Synta:-.. Test

I. Anaphora Tes t A 1. Alice and June took a bath. June washed herse If.

question: Who washed who') answer: washed

2. Mary and Ann are going ta a birthday party. Mary dressed her. 3. John \Vent hunt i ng wi th Bob. By accident, John ki lIed hi m. 4. June and May are friends. May tells a lot ahout herself.

II. Translation 1. John thinks that Bill knows Mary. 2. Tom knows that Paul said that Bob liked cats. 3. !\1ary told Ann ta stand up. 4. Bill toid June not to watch n'. 5. Alice says that John wants Ann ta go ta school. 6. Tom thinks that Bob asked Mary not to eat apples. 7. June gave Paul a birthday present.

Ill. Vocabulary hi t, introduce. understand. hate. remember. believe. blame, picture. talk. show.

IV. Anaphora Test B 1. Mary went hunting with Bob.

a. By accident. Mary killed him. b. By accident, Mary killed her.

2. John and Alice are gai ng ta a party. a. Alice dressed herself. b. Al i ce dressed himself.

3. Bill often fights with Ann. a. Bill al ways hurts him. b. Bill a 1 ways hurts her.

4. May and Paul took a bath. a. Paul washed himself. b. Paul washed herself.

--104-

1

1

,

APPENDIX B: CG Test

Type A: t wo-clause tensed sentence 1. John said that Bill hit himself. (1) 2. June says that Al ire ~~~~!':~t~l}_~~ hersel f. (6) 3. Tom thinks that John hates himself. (11) 4. Ann remembers that Mary iT1tr<?~~_c_ed herself.( 16) 5. Bob knows that Paul blames himself. (21)

Type B: three-clause tensed sentence 1. Alice knows that May thinks that June hi t herse} f. ( 4)

2. Paul thinks that Bob believes that John understands himself.( 19} - - - - ~ - - - - - . - ~ - . - - --

3. May says that Ann knows that Alice hate;:; herself. 4. Bill believes that Tom said that Paul introduced himself. 5. Mary remembers that June said that Alice blamed herself.

Type C: b\o-c1ause infini ti val sentence 1. John toid Bob not ta hit himself. (25)

2. June wants May to ~1}_~~r_~_~~I}_~ herself. (2) 3. Bob wants Tom not to hate himself. (15) 4. Mary asked Ann to ~~t_t:'~~_l:l~~_ herself. (20) 5. May asks Alice not ta ~!~~E: herself. (0)

Type D: three-clause infini tl val sentence 1. June remembers that Alice asked May oot to ~_~t herself. (12) 2. John thinks that Bill wants Torn to understand himself.( 17) w··._._ •• __ • ___ _

3. Ann knows that Mary told June not to h~t_~ herself. (22) 4. Tom says that Paul told Bob ta il}~t:'"q~~_ç~ himself. (5) 5. Bill believes that John wants Paul not to blame himself. (7)

Type E: one-clause (subject-obj~ct) sentence 1. Tom showed Bill a picture of himself. (3) 2. Ann gave Mary a pi rture of herse If. (8) 3. Bob talked ta Paul about himself. (3) 4. June talked ta Alice about herse!f. (18) ----5. John asked Tom about himself. (23)

(24) ( 9)

04}

*The arder of the sentences appeared in the test is gi ven in parentheses.

-105-

APPENDIX C: lndividual Responses

1 lndicated below: subject #, grade, sex, age, responses for Types A-E in order. For Types A, C, and E. choices are: a. NPl b. NP2 c. NP3 d. someone eise e. don't know

Type A [NPl verb [NP2 verb :.sel f] J

Type C [NP 1 verb NP2 (PRO ta verb ~sel f J ] Type E (NPl verb NP2 ~self J

--

For Types B and D, choices are: a. NP1 b. NP2 c. NP3 d. NP 1/2 e. NP2/3 f. NP1/3 g. NPl/2/3 h. someone eise i. don't know

Type B [NPl verb {NP2 verb [NP3 verb-self]]] Type D [NP1 verb [NP2 verb NP3 [PRO ta verb -self]]]

(Type A) (Type B) ( Type C) (Type D) (Type E) S# Gd Sex Age 1234S 1234') 1 2 3 4 5 123413 1 2 3 4 5

._.---- ~----~~--"~----- ----~-- - ---- ---QI 10 F 15 b abc b cac f c a a a a a b b c b a a a a a a 02 10 M 15 b b b b b c bec c a a a a a bbbbb a a a a a 03 10 F 16 c a a b a i a b i b a a a a a a c b b a a a a a a 04 10 F 15 b a b b b c b b c b a a a b a b b c b b a a a a il

05 10 F 15 b b b b 1 ccc c / b b b b / ccccl a a a a a 06 10 M 15 bbbbb ccc c c b b b b b ccc c c b b a a a 07 10 F 15 bbbbb ccc c c b b b b b ccc c c a b a a b 08 10 F 15 bbbbb ccc c c b b b b a ccc c c b a a a b 09 10 F 15 b b b b b ccc c c c t;l c b c b bec e a a a a b 10 10 M 16 c b b b c c g a c g a c cab g e g c i a a a a a 1110 F 15 a b a b b b b bec a a b b b c b bec a b a a b 12 10 F 15 b b b b b c b b c e b b b b e c e abc a a a a a 1310 F 15 b c b b b e e e c c b c b b b cee c c a e a a b

14 11 F 17 b b a b b cbace b b b a b c b b b b a a a a b 15 11 M 17 b b a a b c cl g c a b b a a a b b bec a a b a a 16 11 M 16 bbbbb e cee c bbbbb ceccc a a a a a 17 11 F 16 bbaba e b b c b abbbb bbccc bbaab 18 11 F 17 bbbbb e b bec b b b b b ecbee a a a a b 19 11 M 16 b b a b b cee e e a a a a a b b b c b a a a a b 20 11 M 16 abbbb cee e c b b b b a bec b c a a a a b 21 11 F 16 b b b a a e b b b e a b a a a b c b b b a a a a a 22 11 F 16 b b b b b e cee c b b b b b cee c e a a a a a 23 11 M 17 a a a a a d d cl d cl a a a b a b a a a a a a a a a 24 11 F 16 a cab a a b a e b a c a e a b b e b b a c a a b 25 11 F 16 b b b b b ccc c c b b b b b ccc c c bbdbb 26 II F 17 b b b b / e a e e / a b b a / bec b / a a b b a 27 11 M 16 b a a b b e b b b c b a a a b c b b b c a b b a a

--- 1 0 6

" (Type A) (Type B) (Type C) (Type D) ( T ipe E) S# Gd Sex Age 1 2 3 4 5 1 2 3 4 ;:; 1 2 3 4 5 1 :2 3 4 5 1 :2 3 4 5 28 12 F 17 b c b b c cee c c cac b c e b e e e a c a a c 29 12 F 18 b a a b a c bac c a a a a a b b b b c a a a a a 30 12 F 17 b b b b b ccc r r b h b b b bec c (" a a a a a 31 12 F 17 b b b b b c b bec b b a b b ccc c c a a a a h 32 12 F 17 b b b b b c bec c b b b b b ccc c c b b a a b 33 12 F 17 bbabb cee e c b a a b a e b b e e b b a a c 34 12 M b b b b a c b g c b a a a a a b b b b b a a a a a 35 12 M 17 b a b b a b bec c a a b b b b b c b c a a a a a 36 12 F 17 b cab a b bec c a a abc b b b c a a c a a b 37 12 F 17 b b b b a c b b c b a b a a b b b bac c a a a b 38 12 F 17 b a b b a c bec a c a a b b c b bec a b a a a 39 12 M 18 c b b b b c bec c b a a b e bec c c vbaaa 40 12 M 17 b b b b b ccc c c b b b b b ccc c c a a a a b 41 12 F 18 b c b b b c c f c c b a b a b ccc b c a a a a a 42 12 M 17 b b b b b ccc c c a a a a a b b b b b a b a a a 43 12 M 17 c a a b a g bec a a a b b b f c d c f b a a c b 44 12 F 17 h h b b b ccc c c h a h h h ccc c c a a a a a 45 12 M 17 b b b b / ccc c / b b b b / ccc b / a a a a a

{ 46 13 F 18 b b b b b e ccc c ccc b a e e cee a a a a a 47 13 F 19 b b b b b c b b b b b a b b b ccc c c a a a a b 48 13 F 18 b b a b b b bec c b a b b b bec c c a b a a b 49 13 F 19 b b b b b ccc c c b b b b b ccc c c a b a a a 50 13 F 19 b b b b b ccc c c b h b b b ccc c c a a a a a 51 13 F 19 a b a a a cab c c a a a c a b b b b a c b a a b 52 13 F 19 a b a b a c b b c b b b a a b b cab c a b a a b 53 13 F 18 bbbbb c cl ccc b b b b b ccc c c a a a a b 54 13 F 18 b b c b b ccc c c cac b c b bec c a bac b 55 13 F 19 b b a b b c bec c a a abc b b b c e a b a a b 56 13 F 18 b b b b b ccc c c b a b b b c e ccc a b a a a 57 13 F 19 hbc:bc ccc c c a a b b b c b e b b a c a a b 58 13 F 18 b b c b b b b b c b a cab a b e bec acbab 59 13 F 18 babbb c bec b bbabb c bec c a a a a b 60 13 F 19 bbaab e c bec b b bac bec c c a b a a c 61 13 F 19 a b b b b e ccc c a a a b b b bec c aaaaa 62 13 F 19 bec b b cee c e abc b c cee e c accca 63 13 F 19 b b b b b cl b b c c b a a b b b b b cl c acaaa 64 13 F 18 b a abc a cl cl c c b b b b c f b bec c b a a b 65 13 F 18 b b b b / ccc c 1 b b b b / ccc c / b b a a b

t

-} 07-

1 Adjémian. C. 1976. On the nature of intcrlanguage systems. Language

Learning 26:297-320.

Baker, C. L. 1979. Syntactic theory and the projection problem.

Linguistlc lnquiry 10:533-581.

Baker. M. C. 1 g88. 1 llcorpor~i!-'2Il_~_LTheory of Grammatical Function

Changing. Chicago; the Uni versi ty of Chicago Press.

Berwick. R. 1985. Ihe Acquisi tian of Syntactic Knowledge. Cambridge,

MA; MI T Press.

Bley-Vroman. R. ~n press. The logicrt! problem of second language

learning. In Gass. S. and Schachter. J. (eds.). Linguistic

Perspecti ves on Second Larlguage Acq~isi tL~f!. Cambridge;

Cambridge Uni versi ty Press.

Brown. R. and Hanlon. C. 1970. Derivational complexity and the arder

of acquisition ln child speech. In Hayes. J.R. (ed.), Cognition

and the Development of Lan$uage. New York; Wiley.

Chien. Y. and Wexler K. 1987. Children's acquisi tian of the

locality condition for reflexives and pronouns. Papers and

Reports on Child Language Develo_pment. Stanford University 25.

Chomsky. N. 1981a. Lectures on Government and Bi ndi ng. Dordrecht;

Foris.

Chomsky. N. 1981b. Principles and parameters in syntactic theory. In

Hornstein. N. and Llghtfoot. D. (eds.), Explanation ln

Linguistics: the logical problem of language acquisition.

London; Longman.

Chomsky, N. 1986a. Knowledge of Language: Its Nature, Origin. and

Use. New York; Praeger Publishers.

Chomsky, N. 1986b. Barriers. Cambridge. MA; MIT Press.

Chomsky. N. 1988. Language and Problems of Knowledge. Cambridge, MA;

MIT Press.

Clahsen, H. and Muysken. P. 1986. The availabili ty of universal

grammar to adult and child learners: a study of the acquisition

of German ward arder. ~~~~~Q __ L_an('''-~~~.e __ ~esea!_~h 2 :93-119.

1 0 8 -

t

1

Cook, v. 1985. Chomsky's uni versaI grammar and second language

acquisi t ion. Applied Li nguistics 6 :2-18. Crain, S. and Nakayama, M. 1987. Structure dependence in grammar

formation. language 63:522-543.

Deutsch, W., Kaster, C. and Koster, J. 1986. What can we learn from

children's errors in understandi ng anaphora? li nguistics 24:

203-225. du Plessis, J., Solin, D., Travis, L. and Whi te, L. 1987. UG or not

UG, that is the question: a reply to Clahsen and Muysken.

Second language Research 3:56-75.

Ellis, R. 1986. Understanding Second Language Acquisition. Oxford;

Oxford University Press.

Felix, S. 1988. UG-generated knowledge in adul t second language

acquisition. In Flynn, S. and D'Neil. W. (eds.), Linguistic

Theory in Second language Acquisition. Dordrecht; Kluwer

Academ i c Publ i shers.

Finer, D. and Broselow, E. 1986. Second language acquisi tion of

reflexi ve-bindi ng. NElS 16. Uni versi ty of Massachusetts at

Amherst; Graduate Linguistics Students Association.

Finer, D. and Broselow, E. 1989. Binding parameters in second

1 anguage acquisi t ion. Paper presented at the Second language

Research Forum, UClA, Las Angeles, Feb. 1989. Flynn, S. 1987. A Parameter-Setting Model of l2 Acquisition.

Dordrecht; Reidel.

Fukui, N. 1984. Studies on Japanese anaphora 1: the adjunct subject

hypothesis and zibun. Unpublished ms., MIT.

Fukui. N. 1986. A Theory of Category Projection and Its

Applications. Unpublished doctoral dissertation, MIT.

Gleitman. L., Newport. E. and Gleitman. H. 1984. The current status

of the motherese hypothesis. Journal of Child language 11: 43-79 Hornstein, N. and Lightfoot, D. (eds.) 1981. Explanation in

Linguistics: the logical probl em of language acquisi tion.

London: Longman.

-} 09-

1 Hirsh-Pasek, K.. Treiman. R. and Schneiderman, M. 1984. Brown and

Hanion revisi ted: mothers' sensi tivi ty to ungrammatical forms.

Journal of Child Language Il :81-88.

InDue. K. 1976a. Reflexivization: an interpretive approach. In Shi ba tan i. M. ( ed. ) . Syn tax and Semant i cs '; : Japanese

Generative Grammar. New York; Academie Press.

Inoue. K. 1976b. Henkei Bunpo-to Nihongo. 'Transformational Grammar

and Japanese', Tokyo; Taishukan.

Jakubowicz, C. 1984. On markedness and binding principles. NELS 14.

Uni versi ty of Massachusetts at Amherst; Graduate Linguistics

Students Association.

Kitagawa, Y. 1986. Subjects in .Japanese and English. Unpublished

doctoral dissertation. University of Massachusetts.

Kuno, S. 1973a. lhe Structure of Japanese Language. Cambridge, MA;

MIT Press.

Kuno, S. 1973b. Nihon Bunpo Kenkyu. 'Japanese Grammar' Tokyo;

Tai shukan. Kuno, S. 1978. Danwa-no Bunpo . Discourse Grammar' Tokyo; Taisyukan.

Hirsh-Pasek, K., Treiman, R. and Schneiderman, M. 1984. Brown and

Hanlon revisited: mothers' sensitivity to ungrammatical forms.

Journal of Child Languag~ 11 :81 -88.

Hornstein, N. and Lightfoot, D. (eds.) 1981. Explanation ln

Linguistics: The LogicJl Problem_qLLanguage Acquisition.

London; Longman. Huang, J. and Tang, J. 1988. On the nature of the long-distance

reflexive ln Chinese. Paper presented at NELS. Cornell

Uni versity, November 1988.

Lust, B. (ed.) 1986. Studies in the Acquisition of Anaphora. Vol 1:

Defining the Constraints. Dordrecht; Reidel.

Lust. B. and Mangione, L. 1984. The principal branching direction

parameter in first language acquisi tion of anaphora. NELS 13. University of Massachusetts at Amherst; Graduate Linguistics

Students Associati on.

-110--

1

t

t

Manzini. R. and Wexler, K. 1987. Parameters, Binding Theory, and

learnabili ty. Linguistic lnquiry 18: 413-444.

Mazurkerwich, I. 1988. The Acquisition of infinitive and gerund

complements by second language learners. In Flynn, S. and

Q'Neil. W. {eds.}, Linguistic Theory in Second Language

Acquisition. Dordrecht; Kluwer Academie Publishers.

Narisawa, M. 1988. The effect of the LIon the Subset Principle.

Unpublished ms., McGill Uni versi ty.

O'Grady, W., Suzuki -Wei. Y. and Sook, W.C. 1986. Directionality

preferences in the interpretation of anaphora: data from Korean

and Japanese. Journal of Child language 13:409-420.

Otsu, Y. 1981. Un i versai Grammar and Syntactic Development in

Chidren: Toward a Theory of Syntactic Development. Unpublished

doctoral dissertation, MIT.

Otsu, Y. (ed.) 1987. Kotobakara mita kokoro [Generative Grammar and

Cognitive Science]. Tokyo; Tokyo University Press.

Otsu, Y. and K. Naoi. 1986. Eigo-gakusyu Project Chosa -Hokoku

· Report on the Engl i sh learni ng Project: ms., Tokyo.

Otsu, Y., Van Riemsdijk, H., Inoue, K., Kamio, A. and Kawasaki. N.

(eds.), Studies in Generative Grammar and language Acquisition.

Tokyo; International Christian University.

Pica, P. 1987. On the nature of the reflexivization cycle. NELS 17.

Pinker, S. 1987. The Bootstrapping problem in language acquisition.

In MacWhinney, B. (ed.), Mechanisms of Language Acquisition.

Hillsdale, NJ; Lawrence Erlbaum.

Phi nney, M. 1987. The pro-drop parame ter 1 n second 1 anguage

acquisi tion. 1 n Roeper, T. and Williams, E. (eds.), Parameter

Setting. Dordrecht; Reidel.

Ri tchie, W. 1978. The right roof constraint in adul t-acquired

language. In Ritchie, W. (ed.), Second language Acquisi tion

Research: Issues and Implications. New York; Academie Press.

Schachter, J. 1988a. Second language acquisi tion and i ts relation­

ship ta Universal Grammar. Applied LingYlstics 9:219-235.

-1 1 1-

1 Schachter. J. 1988b. On the issue of completeness in second language

acquisi tion. Paper presented at the Boston Uni versi ty

Conference on Language Development. Boston. Oct. 1988.

Selinker. L. 1972. Interlanguage. IRAL 10:209-231.

Thomas, M. 1988. In Press. The interpretation of EngI ish reflexive

pronouns by non-native speakers. Studies in Second Language

Research.

van Riemsdijk, H. and Williams, E. 1986. Introduction ta the Theory

of Grammar. Cambridge, MA; MIT press.

WexIer, K. and Manzini, R. 1987. Pararneters and Learnability in

Binding Theory. In Rotper. T. and Williams, E. (eds.),

Parame ter Setting. Dordrecht.; Reidel.

Wexler. K. and Chien. Y. 1935. The development of lexical anaphors

and pronouns. Pa pers and Reports on Chi Id Language Development

24:138-149. Stanford University.

Whi te, L. 1985a. The acquisi tion of parameterized grammars:

subjacency in second language acquisi tian. Second Language

Research. 1:1-17.

Whi te, L. 1985b. Is there a logica'l problem of second language

acquisi tian? TESL Canada 2.2 :29 -/11.

White, L. 1986a. Implications of parameteric variation for adult

second language acquisition: an investigation of the 'pro-drop'

parame ter. In Cook, V. (ed.), ~xperimental A~proaches to Second

Language Acquisi tian. Oxford; Pergamon Press.

White, L. 1986b. Markedness and parameter setting: sorne implications

for a theory of adult second language acquisi tion. In Eckman,

F., Moravscik, E. and W i rth, J. (cds.), Markedness. New York;

Plenum Press.

Whi te, L. 1987a. Agai nst comprehensible input: the input hypothesis

and the development of L2 competence. ~pplied Linguistics 8:

95-110.

White, L. 1987b. Markedness and second language acquisition: the

t ransf er. ~~.uAi.e~ _.i n ._~_~.c.Qnd Langua~e_ .. ~cQ,u!si.ti of! 9 :261-286.

-. 1 1 2 -

Whi te, L. 1987c. Uni versaI Grammar. Is it just a new name for oid

problems? Paper presented at Second Language Research Forum,

use, Los Angeles, Feb. 1987.

Whi te, L. 1988. Island effects in second language acquisi tion. In

Flynn, S. and Q'Neil. W. (eds.), Linguistic Theory in Second

Language Acquisïtion. Dordrecht; Kluwer Academic Publishers.

Whi te, L. In press. The principle of adjacency in second language

acquisi tion: do L2 learners observe the subset principle? In

Gass, S. and Schachter, J. (eds.), Linguistic Perspectives on

Second Language Acquisi tion. Cambridge; Cambridge Uni versi ty

Press.

White, L. In preparation. Universal Grammar and Second Language

Acquisi non.

Yang, O.-W. 1988. An Extended Binding Theory. Paper presented at

NELS. Corne Il University, November 1988.

ZobI. H. 1988. Configurationali ty and the subset principle: the

acquisi tion of V' by Japanese learners of English. In

Pankhurst, J., Sharwood Sm i th, M. and Van Buren, P. ( eds.),

Learnabi li ty and Second Languages: A Book of Readings.

Dordrecht; Foris.

-11 3-