digitool.library.mcgill.cadigitool.library.mcgill.ca/thesisfile55607.pdf · 1 1 l , Abstract...
Transcript of digitool.library.mcgill.cadigitool.library.mcgill.ca/thesisfile55607.pdf · 1 1 l , Abstract...
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-
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-
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'Œ gradt;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-